BEGIN:VCALENDAR VERSION:2.0 METHOD:PUBLISH X-WR-CALNAME:IL-Technion-CS-events PRODID:-//tzurl.org//NONSGML Olson 2010k//EN BEGIN:VTIMEZONE TZID:Asia/Jerusalem BEGIN:STANDARD DTSTART:19500910T020000 TZOFFSETFROM:+0300 TZOFFSETTO:+0200 RRULE:FREQ=YEARLY;BYMINUTE=0;BYHOUR=2;BYDAY=2SU;BYMONTH=9 END:STANDARD BEGIN:DAYLIGHT DTSTART:19500331T020000 TZOFFSETFROM:+0200 TZOFFSETTO:+0300 RRULE:FREQ=YEARLY;BYMINUTE=0;BYHOUR=2;BYDAY=-1FR;BYMONTH=3 END:DAYLIGHT END:VTIMEZONE BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100104T143000 DTEND;TZID="Asia/Jerusalem":20100104T163000 DTSTAMP;TZID="Asia/Jerusalem":20100104T143000 FREEBUSY;FBTYPE=BUSY:20100104T143000/20100104T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Ofer M. Shir about Derandomized Search for Experimental Optimization at 2010-01-04 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040007990 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100104T183000 DTEND;TZID="Asia/Jerusalem":20100104T203000 DTSTAMP;TZID="Asia/Jerusalem":20100104T183000 FREEBUSY;FBTYPE=BUSY:20100104T183000/20100104T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Raz Ben Yehuda (Ms. c student in the Open University) about Haifux, Haifa Linux Club: Offline Scheduler at 2010-01-04 18:30:00 DESCRIPTION;LANGUAGE=en-US:OFFSCHED is a platform aimed to assign an assig nment to an offloaded processor. An offloaded processor is a processor t hat is hot un-plugged from the operating system. In today's computer wor ld, we find that most processors have several embedded cores and hyper -threading. Most programmers do not really use these powerful features a nd let the operating system do the work. At most, a programmer will boun d an application to a certain processor or assign an interrupt to a diff erent processor.

At the end, we get a system busy in maint aining tasks across processors, balancing interrupts, flushing TLBs and DTLBs using atomic operations even when not needed and worst of all, spi n locks across processors in vein; and the more processors the merrier. I argue that in some cases, part of this behavior is due to fact the mul tiple core operating system is not service oriented but a system oriente d. There is no easy way to assign a processor to do a distinct service, undisturbed, accurate, and fast as long as the processor is an active pa rt of an operating system and still be a part of most of the operating s ystem address space.

The purpose of the OFFSCHED is to cre ate a platform for services. For example, assume a system is being attac ked; the Linux operating system will generate endless number of interrup ts and/or softirqs to analyze the traffic and throw out bad packets. Thi s is on the expense of good packets. Have you ever tried to ssh an attac ked machine? Who protects the operating system? What if we can simply do the packet analysis outside the operating system without being interrup ted? Why not assign a core to do only fire-walling? Or just routing? Des ign a new type of Real Time system? Maybe assign it as an ultra accurate timer? Create a delaying service that does not just spin? Offload a TCP stack? Perhaps a new type of a locking scheme? New type bottom-halves? Debug a running kernel through an offloaded processor? Maybe assign a GP U to do other things than just graphics? Amdahl Law teaches us that line ar speed-up is not very feasible, so why not spare a processor to do c ertain tasks better? Technologically speaking, I am referring to the Lin ux kernel ability to virtually hot unplug a (SMT) processor; but instead of letting it wonder in endless "halts", assign it a service. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec441040008180 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100105T143000 DTEND;TZID="Asia/Jerusalem":20100105T163000 DTSTAMP;TZID="Asia/Jerusalem":20100105T143000 FREEBUSY;FBTYPE=BUSY:20100105T143000/20100105T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Niv Buchbinder about The Randomi zed k-Server Conjecture (Online Algorithms meet Linear Programming) at 201 0-01-05 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040007940 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100105T143000 DTEND;TZID="Asia/Jerusalem":20100105T163000 DTSTAMP;TZID="Asia/Jerusalem":20100105T143000 FREEBUSY;FBTYPE=BUSY:20100105T143000/20100105T163000 SUMMARY;LANGUAGE=en-US:The msc talk by Javier Turek about On MMSE and MAP Denoising Under Sparse Representation Modeling Over a Unitary Dictionary a t 2010-01-05 14:30:00 DESCRIPTION;LANGUAGE=en-US:Among the many ways to model signals, a recent approach that draws considerable attention is sparse representation model ing. In this model, the signal is assumed to be generated as a random lin ear combination of a few atoms from a pre-specified dictionary. In this work we analyze two Bayesian denoising algorithms -- the Maximum-Aposteri ori Probability (MAP) and the Minimum-Mean-Squared-Error (MMSE) estimator s, under the assumption that the dictionary is unitary. It is well known that both these estimators lead to a scalar shrinkage on the transformed coefficients, albeit with a different response curve. We start by deriv ing closed-form expressions for these shrinkage curves and then analyze t heir performance. Upper bounds on the MAP and the MMSE estimation errors are derived. We tie these to the error obtained by a so-called oracle est imator, where the support is given, establishing a worst-case gain-factor between the MAP/MMSE estimation errors and the oracle's performance. The se denoising algorithms are demonstrated on synthetic signals and on tru e data (images). ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400012060 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100106T133000 DTEND;TZID="Asia/Jerusalem":20100106T153000 DTSTAMP;TZID="Asia/Jerusalem":20100106T133000 FREEBUSY;FBTYPE=BUSY:20100106T133000/20100106T153000 SUMMARY;LANGUAGE=en-US:The Combinatorics Semin talk by Shachar Lovett (We izmann Institute) about Combinatorics Seminar: On Equivalence of Polynomi al Conjectures in Additive Combinatorics at 2010-01-06 13:30:00 DESCRIPTION;LANGUAGE=en-US:We will discuss two important conjectures in ad ditive combinatorics. The first one is the polynomial Freiman-Rusza conj ecture, which relates to the structure of sets with small doubling. The second is the inverse Gowers conjecture for $U^3$, which relates to func tions which locally look like quadratics. In both conjectures a weak for m, with exponential decay of parameters is known, and a strong form with only a polynomial decay of parameters is conjectured.

We w ill show that the two conjectures are in fact equivalent. This was also discovered independently by Green and Tao ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Amado 719 UID:111ec441040008230 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100107T143000 DTEND;TZID="Asia/Jerusalem":20100107T163000 DTSTAMP;TZID="Asia/Jerusalem":20100107T143000 FREEBUSY;FBTYPE=BUSY:20100107T143000/20100107T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Eran Tromer about Side Channels and their Mitigation in Cloud Computing Security at 2010-01-07 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008010 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100112T113000 DTEND;TZID="Asia/Jerusalem":20100112T133000 DTSTAMP;TZID="Asia/Jerusalem":20100112T113000 FREEBUSY;FBTYPE=BUSY:20100112T113000/20100112T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Tammy Riklin-Raviv (Medical Vision Group CSAIL, MIT & Surgical Planning Laboratory, Harvard Medical Sc hool) about Pixel Club Seminar: Segmentation of Image Ensembles via Latent Atlases at 2010-01-12 11:30:00 DESCRIPTION;LANGUAGE=en-US:The images acquired via medical imaging modalit ies are frequently subject to low signal-to-noise ratio, bias field and pa rtial volume effects. These artifacts, together with the naturally low con trast between image intensities of some neighboring structures, make the e xtraction of regions of interest (ROIs) in clinical images a challenging p roblem.

Probabilistic atlases, typically generated from compr ehensive sets of manually labeled examples, facilitate the analysis by pro viding statistical priors for tissue classification and structure segment ation. However, the limited availability of

training examples that are compatible with the images to be segmented renders the atlas-bas ed approaches impractical in many cases.

In the talk I will p resent a generative model for joint segmentation of corresponding regions of interest in a collec- tion of aligned images that does not require labe led training data. Instead, the evolving segmentation of the entire image set supports each of the individual segmentations. This is made possible b y iteratively inferring a subset of the model parameters, called the spati al parameters, as part of the joint segmentation processes. These spatial parameters are defined in the image domain and can be viewed as a latent a tlas, that is used as a spatial prior on the tissue labels. Our latent atl as formulation is based on probabilistic principles, but we solve it using partial differential

equations (PDEs) and energy minimizatio n criteria. We evaluate the method successfully for the segmentation of co rtical and subcortical structures within different populations and of brai n tumors in a single-subject multi-modal longitudinal experiment. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec441040008210 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100118T143000 DTEND;TZID="Asia/Jerusalem":20100118T163000 DTSTAMP;TZID="Asia/Jerusalem":20100118T143000 FREEBUSY;FBTYPE=BUSY:20100118T143000/20100118T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Michael Schapira about Internet Routing: Foundations, Challenges and Future Directions at 2010-01-18 14:30 :00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008190 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100118T183000 DTEND;TZID="Asia/Jerusalem":20100118T203000 DTSTAMP;TZID="Asia/Jerusalem":20100118T183000 FREEBUSY;FBTYPE=BUSY:20100118T183000/20100118T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Yaron Dishon (Socio logy Dept. student, Haifa University) about Haifux, Haifa Linux Club: The FOSS Community as a Social Phenomenon at 2010-01-18 18:30:00 DESCRIPTION;LANGUAGE=en-US:What kind of community is it? Who are its membe rs? What are its boundaries?

מחקר זה מנסה להבי מהם התהליכים שהובילו להתכוננות קהילת ה וד הפתוח מתוך קבוצת האנשים הפעילים בפר ויקטים השונים. על אף שקהילת הקוד הפתוח נח קרה רבות מנקודת מבט טכנולוגית וארגונית , המחקר הסוציולוגי, היכול לתרום רבות להבנ ת התהליכים החברתיים המתרחשים במסגרת הק הילה, נמצא בחסר. על מנת לעקוב אחר תהליכים שהובילו לכינון הקהילה מצייר המחקר תמונה אודות תהליכי ההצטרפות של חברים חדשים ב הילה. בסדרה של ראיונות עם חברים בקהילת ה וד הפתוח הישראלית נמצא כי תהליך ההשתלב ות בקהילה כולל רכישת ידע וערכים, סטאטוס ק הילתי ותחושת שייכות. בניתוח הממצאים וב בט על קהילת הקוד הפתוח כ'קהילת מעשה', מעל המחקר כי בדומה לקהילות מעשה אחרות מתח מים אחרים, קהילת הקוד הפתוח התכוננה 'מעצ ה', כתוצאה מהרצון של חבריה לפיתוח תוכנה והעמקת כישורים אישיים. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec441040008270 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100119T113000 DTEND;TZID="Asia/Jerusalem":20100119T133000 DTSTAMP;TZID="Asia/Jerusalem":20100119T113000 FREEBUSY;FBTYPE=BUSY:20100119T113000/20100119T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Marina Alterman (EE, Technio n) about Pixel Club Seminar: Theory of Multiplexed Fluorescence Unmixing at 2010-01-19 11:30:00 DESCRIPTION;LANGUAGE=en-US:Fluorescence microscopy is a powerful tool in b iology and biomedical sciences. Microscopic specimens usually yield fluore scence intensity images that are dim and thus suffer from low signal-to-no ise ratio (SNR). Moreover, in multispectral imaging of fluorescing specime n, intensities are just a means to obtain information about molecular dist ributions of the materials in the specimen.

Multiplexed sens ing is a way for increasing the signal-to-noise ratio (SNR) of intensity d ata arrays, without increasing acquisition resources such as time. However , as in fluorescence, these arrays themselves are often not the ultimate g oal of a system. For example, spectral reflectance, emission or absorption distributions stem from an underlying mixture of materials. Systems thus try to infer concentrations of these underlying mixed components.
The process of inverting mixtures is termed unmixing. It is central i n many problems. We incorporate the mixing/unmixing process together with detector noise characteristics explicitly into the optimization of multipl exing codes. This enables optimal recovery of underlying components (mater ials). In the absence of this integrated optimization, multiplexed imaging can even harm the quality of unmixing. Moreover, by directly defining the goal of data acquisition to be recovery of components (materials) instead of intensity/reflectance arrays, the acquisition becomes more efficient. We thus develop a theory for multiplexed sensing, in which the end task is linear unmixing. This yields significant generalizations of multiplexing theory.

* M.Sc. research under supervision of Prof. Yoav Shec hner ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec441040008240 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100119T143000 DTEND;TZID="Asia/Jerusalem":20100119T163000 DTSTAMP;TZID="Asia/Jerusalem":20100119T143000 FREEBUSY;FBTYPE=BUSY:20100119T143000/20100119T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Joe Halpern about Beyond Nash Eq uilibrium: Solution Concepts for the 21st Century at 2010-01-19 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008250 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100121T143000 DTEND;TZID="Asia/Jerusalem":20100121T163000 DTSTAMP;TZID="Asia/Jerusalem":20100121T143000 FREEBUSY;FBTYPE=BUSY:20100121T143000/20100121T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Iftach Ilan Haitner about A Para llel Repetition Theorem for Any Cryptographic Protocol at 2010-01-21 14:30 :00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008170 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100124T143000 DTEND;TZID="Asia/Jerusalem":20100124T163000 DTSTAMP;TZID="Asia/Jerusalem":20100124T143000 FREEBUSY;FBTYPE=BUSY:20100124T143000/20100124T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Naoya Maruyama about Acceleratin g the TSUBAME Supercomputer with Graphics Processing Units and its Implica tions to Systems Research at 2010-01-24 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008150 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100126T113000 DTEND;TZID="Asia/Jerusalem":20100126T133000 DTSTAMP;TZID="Asia/Jerusalem":20100126T113000 FREEBUSY;FBTYPE=BUSY:20100126T113000/20100126T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Yael Pritch (School of Compu ter Science and Engineering ,The Hebrew University of Jerusalem) about Pix el Club Seminar: Shift-Map Image Editing at 2010-01-26 11:30:00 DESCRIPTION;LANGUAGE=en-US:Geometric rearrangement of images includes oper ations such as image retargeting, object removal, or object rearrangement. Each such operation can be characterized by a shift-map: the relative shi ft of every pixel in the output image from its source in an input image. W e describe a new representation of these operations as an optimal graph la beling, where the shift-map represents the selected label for each output pixel. Two terms are used in computing the optimal shift-map: (i) A dat a term which indicates constraints such as the change in image size, objec t rearrangement, a possible saliency map, etc. (ii) A smoothness term, minimizing the new discontinuities in the output image caused by discontin uities in the shift-map. This graph labeling problem can be solved usin g graph cuts. Since the optimization is global and discrete, it outperform s state of the art methods in most cases. Efficient hierarchical solutions for graph-cuts are presented, and operations on 1M images can take only a few seconds.

*Joint work with Eitam Kav-Venaki, and Shmuel P eleg ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec441040008260 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100128T143000 DTEND;TZID="Asia/Jerusalem":20100128T163000 DTSTAMP;TZID="Asia/Jerusalem":20100128T143000 FREEBUSY;FBTYPE=BUSY:20100128T143000/20100128T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Tamir Tuller about Computational Modeling and Algorithms in Molecular Evolution and Gene Translation at 20 10-01-28 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008200 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100128T160000 DTEND;TZID="Asia/Jerusalem":20100128T180000 DTSTAMP;TZID="Asia/Jerusalem":20100128T160000 FREEBUSY;FBTYPE=BUSY:20100128T160000/20100128T180000 SUMMARY;LANGUAGE=en-US:The msc talk by Artyom Sharov about Coding Techniqu es for Multidimensional Constrained Channels at 2010-01-28 16:00:00 DESCRIPTION;LANGUAGE=en-US:Constrained systems are models for describing t he read-write requirements of secondary storage systems, such as magnetic disks and optical devices. Proposals for new storage systems, such as holographic storage, and for better exploitation of current optical syste ms, have raised the interest in 2-D and 3-D constraints to model the read -write requirements of the emerging storage media. In this talk, we pre sent two new coding techniques for 2-D constraints. The first technique i s a variable-rate coding scheme that is based on tiling, namely, on perio dic partitioning of the 2-D plane. In the simplest scenario, there are tw o classes of tiles, referred to as white tiles and black tiles. The shape s and the positioning of the tiles are chosen so that the encoder can fir st assign a value to each white tile independently of the values assigned to the other white tiles, and then assign values to each black tiles de pending only on neighboring already-assigned white tiles. The rate of the encoder is easily calculated, thereby yielding a lower bound on the capa city of the respective constraint. For certain constraints, our technique is shown to improve on previously-known lower bounds on the capacity of the constraint. We then present generalizations of the basic scenario, su ch as allowing certain dependence among the values assigned to the white tiles. We conclude the talk by presenting a second coding scheme, obtai ned through a transformation of our variable-rate scheme into a fixed-ra te encoder, with virtually no loss in the rate. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400012200 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100201T183000 DTEND;TZID="Asia/Jerusalem":20100201T203000 DTSTAMP;TZID="Asia/Jerusalem":20100201T183000 FREEBUSY;FBTYPE=BUSY:20100201T183000/20100201T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Nadav Har'El (IBM H RL) about Haifux, Haifa Linux Club: Hspell - A Retrospective at 2010-02-0 1 18:30:00 DESCRIPTION;LANGUAGE=en-US:Hspell is a free Hebrew spell checker. It is us ed by most Linux distributions, by free applications such as OpenOffice and Firefox, and even by Google's popular Gmail service. Seven years aft er Hspell's first release, it is a good occasion to look back and see wh at made it successful. We will review Hspell's design, how it was works, and how it was developed. We will ask ourselves what allowed Hspell to be developed quickly, what ensured its quality, and what made it easy to adopt by all those systems and applications. Finally, we will look at what Hspell can do beyond just spell-checking, and speculate which new linguistic capabilities will be needed by the applications of the next d ecade. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec441040008340 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100203T140000 DTEND;TZID="Asia/Jerusalem":20100203T160000 DTSTAMP;TZID="Asia/Jerusalem":20100203T140000 FREEBUSY;FBTYPE=BUSY:20100203T140000/20100203T160000 SUMMARY;LANGUAGE=en-US:The msc talk by Katherine Barabash about Scalable G arbage Collection on Highly Parallel Platforms at 2010-02-03 14:00:00 DESCRIPTION;LANGUAGE=en-US:Computing landscape is changing rapidly in the recent years. On the one hand, the pervasiveness of multiprocessor and m ulticore hardware requires the software to be able to take advantage of the increasingly available parallelism. On the other hand, the growing c omplexity of the modern software application domains makes runtime langu age environments more popular as a major software development tool. In this work, we investigate a question whether a garbage collector, being a n important part of the modern runtime language environment, is able to deal with the potential higher parallelism of tomorrow's hardware platfo rms. We argue that the structure of the application object graph in a ga rbage collected heap can influence the ability of a collector to scale. In particular, certain, sequential in nature, patterns in the object gra ph structure can prevent the tracing garbage collector from scaling the important collection phase -- tracing through the live objects graph. First, we examine the object graphs created by the standard Java benchmark s and describe the idealized trace utilization measure we use to evaluat e applications in terms of their ability to sustain parallel tracing. Next, we present two solutions for alleviating the scalability problems caused by the problematic object graph properties. The first solution lets the system add pointers to the the headers of objects, which artificial ly modify the object-graph shape and make it more scalable. The second s olution is to let additional garbage collection threads run on idle proces sors. We present and analyze the results obtained by evaluating our pro totypes for both solutions. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012070 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100210T110000 DTEND;TZID="Asia/Jerusalem":20100210T130000 DTSTAMP;TZID="Asia/Jerusalem":20100210T110000 FREEBUSY;FBTYPE=BUSY:20100210T110000/20100210T130000 SUMMARY;LANGUAGE=en-US:The phd talk by Mark Silberstein about An online di stributed system for genetic linkage analysis at 2010-02-10 11:00:00 DESCRIPTION;LANGUAGE=en-US:In this talk I will describe the algorithms and mechanisms underlying a distributed system for genetic linkage analysis, called Superlink-online. It is a production online system which serves hu ndreds of geneticists worldwide allowing for faster analysis of genetic da ta via automatic parallelization and execution on thousands of non-dedica ted computers. I will describe the following innovative technologies fo rming the core of this system 1. Practical scheduling and execution of embarrassingly parallel Bags of Tasks in multiple non-dedicated computing environments (SC09). Our approach allows for virtualization of multiple g rids, clouds and volunteer grids as a single computing platform by buildin g an overlay of execution clients over the physical resources; another com ponent is a generic mechanism for dynamic scheduling policies to reduce t he turnaround time in the presence of resource failures and heterogeneity. Our system has executed hundreds of Bags of Tasks with over 9 million job s during 3 months alone; these have been invoked on 25,000 hosts from the local clusters, the Open Science Grid, EGEE, UW Madison pool and Superli nk@Technion community grid. 2. A general technique for designing memo ry-bound algorithms on GPUs through software-managed cache (ICS08). This technique was successfully applied to the probabilistic network inference yielding an order of magnitude performance improvement versus the perform ance without such a cache. Overall we achieved up to three orders of magn itude speedup when executing our GPU-based algorithm versus single CPU per formance. 3. Coarse- and fine-grained parallel algorithms for the inf erence in probabilistic networks on large-scale non-dedicated environmen ts and GPUs. We devised and implemented an algorithm suitable for loosly coupled environments with unreliable resources (American Journal of Human Genetics 2006, HPDC06) and adapted it for heterogeneous GPU-CPU supercomp uter TSUBAME in Tokyo Institute of Technology. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012120 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100210T133000 DTEND;TZID="Asia/Jerusalem":20100210T153000 DTSTAMP;TZID="Asia/Jerusalem":20100210T133000 FREEBUSY;FBTYPE=BUSY:20100210T133000/20100210T153000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Massimo Lauria (Universi tà di Roma) about Theory Seminar: The Strength of Parameterized Tree-like Resolution at 2010-02-10 13:30:00 DESCRIPTION;LANGUAGE=en-US:We examine the proof-theoretic strength of para meterized tree-like resolution---a proof system for the $\co\W[2]$-comp lete set of parameterized tautologies.

Parameterized resolut ion and, moreover, a general framework for parameterized proof complexi ty was introduced by Dantchev, Martin, and Szeider (FOCS'07). In that p aper, Dantchev et al.\ show a complexity gap in parameterized tree-like resolution for propositional formulas arising from translations of fir st-order principles.

Here we pursue a purely combinatoria l approach to obtain lower bounds to the proof size in parameterized tr ee-like resolution. For this we devise a prover-delayer game suitable f or parameterized resolution. By exhibiting good delayer strategies we t hen show lower bounds for the pigeonhole principle as well as the order principle. On the other hand, we demonstrate that parameterized tree-l ike resolution is a very powerful system, as it allows short refutation s of all parameterized contradictions given as bounded-width CNF's. Thu s, a number of principles such as Tseitin tautologies, pebbling contrad ictions, or random 3-CNF's which serve as hard examples for classical r esolution become easy in the parameterized setting.

Joint wo rk with Olaf Beyersdorff and Nicola Galesi. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec441040008380 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100215T143000 DTEND;TZID="Asia/Jerusalem":20100215T163000 DTSTAMP;TZID="Asia/Jerusalem":20100215T143000 FREEBUSY;FBTYPE=BUSY:20100215T143000/20100215T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Gwendal Simon SPECIAL TALK about Optimal Network Locality in Distributed Services at 2010-02-15 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008390 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100215T183000 DTEND;TZID="Asia/Jerusalem":20100215T203000 DTSTAMP;TZID="Asia/Jerusalem":20100215T183000 FREEBUSY;FBTYPE=BUSY:20100215T183000/20100215T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Rami Rosen about Ha ifux, Haifa Linux Club: VoIP in Linux at 2010-02-15 18:30:00 DESCRIPTION;LANGUAGE=en-US:VoIP is an emerging and exciting technology. We will deal with the basics of VoIP protocols and we will discuss some Linux VoIP applications. We will also discuss VOIP with cellular phones (like Android).


VOIP protocols:

- RTP - Real Time protocol.

- RT CP - Real Time control protocol.

 

Codecs (audio and video):

- SIP -the Session Initiation Protocol.

  --Sip Requests and Sip Responses.

  --The INVITE, REGISTER and Byerequests .

 

SIP clients and Sip Proxy servers:

- Open source sip clients:

  --Ekiga, formerl y GnomeMeeting.

&nbs p; --SipDroid (running under Android).

 

Open source sip server (kamailio and opensips, for merly SER).
Voip with Cellular devices (Android)

 

Note: this lecture does not require any previous backgroun d in VoIP.

 

Slides:

voipToHaifux.pdf. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec441040008410 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100216T143000 DTEND;TZID="Asia/Jerusalem":20100216T163000 DTSTAMP;TZID="Asia/Jerusalem":20100216T143000 FREEBUSY;FBTYPE=BUSY:20100216T143000/20100216T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Tal Moran about Cryptography by the People, for the People: How Voting and Cryptography Go Hand-in-Hand at 2010-02-16 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008300 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100217T153000 DTEND;TZID="Asia/Jerusalem":20100217T173000 DTSTAMP;TZID="Asia/Jerusalem":20100217T153000 FREEBUSY;FBTYPE=BUSY:20100217T153000/20100217T173000 SUMMARY;LANGUAGE=en-US:The msc talk by Kolman Vornovitsky about Abstractio ns for devising compact controllers for MDPs at 2010-02-17 15:30:00 DESCRIPTION;LANGUAGE=en-US:Planning a course of action is a key ability fo r intelligent systems. It involves the representation of actions and world models, reasoning about the effects of actions, and techniques for effici ently searching the space of possible plans. Planning under uncertainty is captured by the area of decision-theoretic planning (DTP). In such proble ms, the actions have stochastic effects, and the goal is to devise a polic y of acting with a high expected utility, as opposed to deterministic plan ning which looks for sequence of actions that translates an agent from an initial state to a goal state. The most classical mathematical framework i n use of decision-theoretic planning is this of Markov decision processes (MDP). One of the most interesting insights emerging from AI planning , reasoning under uncertainty, decision analysis and OR is that many DTP p roblems exhibit considerable structure. In particular, the use of variable -based representations to describe problems, as is the typical practice in planning. However most of the solutions for MDP's, i.e. algorithms which compute the optimal policy, assume an explicit state space. We are us ing a technique known in deterministic planning as over-approximating abst ractions to approximately solve the exponential state space MDP problem. V ery roughly, an abstraction is a mapping that reduces the size of the stat e space by contracting several states into one. By making the abstract spa ce small enough, it becomes feasible to perform the standard solutions for explicit state space MDP's on the abstract state space. The ``merge- and-shrink'' methodology introduced by Drager, Finkbeiner & Podelski was a dopted as over-approximations abstraction in the context of deterministic planning by Helmert, Haslum & Hoffmann. We adapt the latter technique for devising compact controllers for MDP's. Additionally we extend the ``merge -and-shrink'' both for MDP's and for deterministic planning with action ab straction techniques. Finally, we provide a clear testbed evaluation for o ur methods, and compare their effectiveness with other state-of-the-art ap proaches. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012090 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100218T143000 DTEND;TZID="Asia/Jerusalem":20100218T163000 DTSTAMP;TZID="Asia/Jerusalem":20100218T143000 FREEBUSY;FBTYPE=BUSY:20100218T143000/20100218T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Ofer Dekel about Learning from M ultiple Teachers at 2010-02-18 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008310 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100223T110000 DTEND;TZID="Asia/Jerusalem":20100223T130000 DTSTAMP;TZID="Asia/Jerusalem":20100223T110000 FREEBUSY;FBTYPE=BUSY:20100223T110000/20100223T130000 SUMMARY;LANGUAGE=en-US:The msc (otherfac) talk by Alexander Raginsky about IE Student M.Sc. Seminar: Optimizing the Size of Software Inspection Team s at 2010-02-23 11:00:00 DESCRIPTION;LANGUAGE=en-US:Code inspection is considered to be an efficien t method for detecting some kinds of faults in software code documents. Ot her kinds of faults are more efficiently detected by other methods such as testing. It has therefore been suggested first to inspect the code and th ereafter test it. The number of inspectors employed in the inspection shou ld be determined to enable detection of most of the faults that inspection detects efficiently. The faults detected by the inspection will correcte d, such that the following testing will operate on code containing by and large only faults of kinds that testing detects more efficiently than insp ection. The number of faults that have not been detected by an inspection may be estimated by the labor intensive faults injection method or by Capt ure recapture metods that seem to have accuracy problems. These estimators do, however, not meet the need to estimate the number of required inspect ors ahead of the inspection. This number may be estimated by an estimator developed by Kantorowitz.

Past experiments suggested t he usefulness of this estimator for inspection of user requirements docume nts. This research encompassed an experimental evaluation of the usefulnes s of th estimator for code inspections. These experiments employed college students, university students, beginner engineers and experienced enginee rs. These experiments suggest that the estimator is also useful for code i nspections. Additional benefits of this estimator are estimates of metrics that quantifies the performance of the employed inspectors. The numerical values of these metrics may be employed for comparing the performance of different kinds of inspectors and in estimating the costs of future inspec tion with similar inspectors. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec441040008440 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100301T183000 DTEND;TZID="Asia/Jerusalem":20100301T203000 DTSTAMP;TZID="Asia/Jerusalem":20100301T183000 FREEBUSY;FBTYPE=BUSY:20100301T183000/20100301T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Michael Shmilov Yon i Limor (yam-design) and Israel David about Haifux, Haifa Linux Club: Dru pal at 2010-03-01 18:30:00 DESCRIPTION;LANGUAGE=en-US:Drupal is an open source content management sys tem that developed into a web application framework written in PHP, it is used to build a wide variety of high-profile web sites, including open sou rce projects like the Linux Journal, online news publishers like The New Y ork Observer , and social networking communities like MTV UK. The talk wil l begin with an overview of Drupal: its capabilities and its usage profile . Some ready-to-use Drupal full-blown installs will be shown as well. The second half of the talk will give us an idea of what it feels like to swit ch to Drupal, after seven years of web site development. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec441040008420 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100303T140000 DTEND;TZID="Asia/Jerusalem":20100303T160000 DTSTAMP;TZID="Asia/Jerusalem":20100303T140000 FREEBUSY;FBTYPE=BUSY:20100303T140000/20100303T160000 SUMMARY;LANGUAGE=en-US:The msc talk by Viacheslav Chernoy about On the Per formance of Dijkstra's 3rd Self-stabilizing Algorithm for Mutual Exclusion and Related Algorithms at 2010-03-03 14:00:00 DESCRIPTION;LANGUAGE=en-US:In a paper of 1974 Dijkstra introduced the noti on of self-stabilizing algorithms and presented three such algorithms for the problem of mutual exclusion on a ring of $n$ processors. The third algorithm is the most interesting of these three but is rather non intuiti ve. In 1986 a proof of its correctness was presented by Dijkstra, but th e question of determining its worst case complexity --- that is, providi ng an upper bound on the number of moves of this algorithm until it stab ilizes --- remained open. In this work we solve this question and prove an upper bound of $3\frac{13}{18} n^2 + O(n)$ for the complexity of thi s algorithm. We also show a lower bound of $1\frac{5}{6} n^2 - O(n)$ for the worst case complexity. For computing the upper bound, we use two tech niques: potential functions and amortized analysis. We also present a ne w-three state self-stabilizing algorithm for mutual exclusion and show a tight bound of $\frac{5}{6} n^2 + O(n)$ for the worst case complexity o f this algorithm. In 1995 Beauquier and Debas presented a similar three- state algorithm, with an upper bound of $5\frac{3}{4}n^2+O(n)$ and a lower bound of $\frac{1}{8}n^2-O(n)$ for its stabilization time. For this alg orithm we prove an upper bound of $1\frac{1}{2}n^2 + O(n)$ and show a lo wer bound of $n^2 - O(n)$. As far as the worst case performance is consi dered, the algorithm of Beauquier and Debas is better than the one of Di jkstra and our algorithm is better than both. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012220 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100304T110000 DTEND;TZID="Asia/Jerusalem":20100304T130000 DTSTAMP;TZID="Asia/Jerusalem":20100304T110000 FREEBUSY;FBTYPE=BUSY:20100304T110000/20100304T130000 SUMMARY;LANGUAGE=en-US:The msc talk by Motty Porat about Construction of S emantic User Interfaces at 2010-03-04 11:00:00 DESCRIPTION;LANGUAGE=en-US:A state of the art Graphical User Interface (GU I) is composed of concrete controls, such as buttons, menus and windows. A semantical GUI, on the other hand, is composed of semantic controls ( Kantorowitz and Lyakas 2005), which are specified by what they do and not by their geometrical and physical properties. A semantic control is thus an abstraction for all the concrete controls that implement the same ac tivity. An example is the "select one of n given choices" semantic contr ol, which may be implemented either by a menu or by a set of n radio but tons. Programming an interactive application with semantic controls is, generally, significantly easier than programming it with the usual concret e controls. The reason is that the programmer need not select and specif y the details of the concrete controls to be employed. The use of s emantic controls requires, however, the availability of code that transl ates the semantic GUI into real-life concrete GUI. In this research we d eveloped the WebSIX framework, in order to experiment with the programming of applications employing semantic GUI's. Programming with semantic con trols whose geometries are not known is quite different from programming with concrete controls. In addition, we looked for solutions that meet some of the many published GUI design patterns. We present the code ar chitecture solutions developed in our research. As expected, the use of abstract semantic controls resulted in code that was quite easy to under stand. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012130 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100307T130000 DTEND;TZID="Asia/Jerusalem":20100307T150000 DTSTAMP;TZID="Asia/Jerusalem":20100307T130000 FREEBUSY;FBTYPE=BUSY:20100307T130000/20100307T150000 SUMMARY;LANGUAGE=en-US:The cggc talk by Roi Poranne (Computer Science, Tec hnion) about CGGC Seminar: 3D Surface Reconstruction Using A Generalized D istance Function at 2010-03-07 13:00:00 DESCRIPTION;LANGUAGE=en-US:3D scanning is the process of acquiring a digit al copy of a physical object. A 3D scanner is used to sample points on the surface of the object (the so-called underlying surface) and acq uire their Cartesian coordinates, after which a surface reconstructi on algorithm is applied to generate a surface based on this sample set. We define a generalized distance function on an unoriented 3D point set and describe how it may be used to reconstruct a surface approximating these points. This distance function is shown to be a Mahalanobis dist ance in a higher-dimensional embedding space of the points, and the res ulting reconstruction algorithm is a natural extension of the classical Radial Basis Function (RBF) approach. Experimental results show the su periority of our reconstruction algorithm to RBF and other methods in a variety of practical scenarios ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008510 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100309T113000 DTEND;TZID="Asia/Jerusalem":20100309T133000 DTSTAMP;TZID="Asia/Jerusalem":20100309T113000 FREEBUSY;FBTYPE=BUSY:20100309T113000/20100309T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Moti Freiman (CS, Hebrew Uni versity of Jerusalm) about Pixel Club Seminar: Shape-constrained Graph Min -cut approach for Medical Image Segmentation at 2010-03-09 11:30:00 DESCRIPTION;LANGUAGE=en-US:Segmentation of organs and vascular structures from clinical Computed Tomography (CT) images is a crucial task in many cl inical applications including diagnosis, patient specific training simulat ions, and intra-operative navigation. The segmentation is a challenging ta sk due to the unclear distinction between the required structure and its surrounding tissue, artifacts in the CT images, and the presence of pathol ogies.

We present a shape constrained graph min-cut approach for the segmentation. Discrete energy functions are defined with respect to f ixed or latent shape models and optimized using the graph min-cut algorith m to obtain an accurate segmentation. Extensive evaluation of our approach for different tasks, including carotid artery bifurcation segmentation, p atient specific modeling of the entire carotid arteries system for simulat ion, abdominal aortic aneurysms segmentation, and kidney and liver segment ation shows that our method is accurate, robust, and practical for clinica l use. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec441040008460 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100309T143000 DTEND;TZID="Asia/Jerusalem":20100309T163000 DTSTAMP;TZID="Asia/Jerusalem":20100309T143000 FREEBUSY;FBTYPE=BUSY:20100309T143000/20100309T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Eldar Fischer about Property Tes ting and its application in databases at 2010-03-09 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008450 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100310T123000 DTEND;TZID="Asia/Jerusalem":20100310T143000 DTSTAMP;TZID="Asia/Jerusalem":20100310T123000 FREEBUSY;FBTYPE=BUSY:20100310T123000/20100310T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Moran Feldman (CS, Techn ion) about Theory Seminar: Non-Preemptive Buffer Management for Latency Se nsitive Packets at 2010-03-10 12:30:00 DESCRIPTION;LANGUAGE=en-US:The delivery of latency sensitive packets is a crucial issue in real time applications of communication networks. Such packets often have a firm deadline and a packet becomes useless if it arri ves after its deadline. The deadline, however, applies only to the packe t's journey through the entire network; individual routers along the pac ket's route face a more flexible deadline.

We consider policies for admitting latency sensitive packets at a router. Each packet is tagged with a value and a packet waiting at a router loses value over time as its probability of arriving at its destination decreases. The router is modeled as a non-preemptive queue, and its objective is to maximize the to tal value of the forwarded packets. When a router receives a packet, it must either accept it (and possibly delay future packets), or reject it immediately. The best policy depends on the set of values that a packet can take. We consider three natural settings: unrestricted model, real-val ued model, where any value above $1$ is allowed, and an integral-valued model.

We obtain the following results. For the unrestricted mod el, we prove that there is no constant competitive ratio algorithm. The real valued model has a randomized 4-competitive algorithm and a matchin g lower bound. We also give for the last model a deterministic lower bound of ~4.236, almost matching the previously known 4.24-competitive algori thm. For the integral-valued model, we show a deterministic 4-competitiv e algorithm, and prove that this is tight even for randomized algorithms. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec441040008550 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100315T143000 DTEND;TZID="Asia/Jerusalem":20100315T163000 DTSTAMP;TZID="Asia/Jerusalem":20100315T143000 FREEBUSY;FBTYPE=BUSY:20100315T143000/20100315T163000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Tzachi Pilpel (Wei zmann Institute of Science) about Bioinformatics Forum: An Universal Adapt ive Translation Efficiency Profile of Proteins at 2010-03-15 14:30:00 DESCRIPTION;LANGUAGE=en-US:Recent years have seen intensive progress in me asuring protein translation. However, the contributions of coding sequen ces to the efficiency of the process remain unclear. Here, we identify a universally conserved profile of translation efficiency along mRNAs c omputed based on adaptation between coding sequences and the tRNA pool. In this profile, the first ~30-50 codons are, on average, translated wit h a low efficiency. Additionally, in eukaryotes, the last ~50 codons sho w the highest efficiency over the full coding sequence. The profile pred icts accurately position-dependent ribosomal density along yeast genes . These data suggest that translation speed, and as a consequence, ribos omal density, are encoded by coding sequences and the tRNA pool. We sugg est that the slow "ramp" at the beginning of mRNAs serves as a late stag e of translation initiation, forming an optimal and robust means to redu ce ribosomal traffic jams, thus minimizing the cost of protein expressio n.

Joint with Tamir Tuller, Asaf Carmi, Sivan Navon et al. In Press in Cell 2010

Host: Zohar Yakhini ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008540 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100315T183000 DTEND;TZID="Asia/Jerusalem":20100315T203000 DTSTAMP;TZID="Asia/Jerusalem":20100315T183000 FREEBUSY;FBTYPE=BUSY:20100315T183000/20100315T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Shimon Panfil (Ph.D .) about Haifux, Haifa Linux Club: Scientific Programming with Modern Fort ran at 2010-03-15 18:30:00 DESCRIPTION;LANGUAGE=en-US:Fortran is the most known language for scientif ic and engineering problems. However growth of popularity of C/C++,Java, Matlab etc has shadowed the development and mere existence of Fortran f rom wider community. Situation may change know. Fortran 2003 standard pr ovides all language features, one expects from modern programming langua ge and gfortran which replaced g77 in gcc starting from version 4.0 impl ements this standard.

Slides are already available. . ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 8 (note different place) UID:111ec441040008530 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100316T113000 DTEND;TZID="Asia/Jerusalem":20100316T133000 DTSTAMP;TZID="Asia/Jerusalem":20100316T113000 FREEBUSY;FBTYPE=BUSY:20100316T113000/20100316T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Oliver Deussen (University o f Konstanz, Germany) about Pixel Club Seminar: Computational Aesthetics - Science or Art? at 2010-03-16 11:30:00 DESCRIPTION;LANGUAGE=en-US:Apart from rendering photorealistic images many works in computer graphics focus on creating illustrative and artistic im ages. Such images can be used in many contexts ranging from CAD to plannin g and technical documentation. Finding mathematical principles and algorit hms for aesthetic configurations is one important aspect for producing goo d illustrations. The talk will give an overview about this fascinating fie ld between art and science, many examples will illustrate how close compua tional outputs come to what artists are able to create. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec441040008470 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100316T143000 DTEND;TZID="Asia/Jerusalem":20100316T163000 DTSTAMP;TZID="Asia/Jerusalem":20100316T143000 FREEBUSY;FBTYPE=BUSY:20100316T143000/20100316T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Igor Razgon about Fixed-paramete r algorithms for graph separation problems at 2010-03-16 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008080 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100317T123000 DTEND;TZID="Asia/Jerusalem":20100317T143000 DTSTAMP;TZID="Asia/Jerusalem":20100317T123000 FREEBUSY;FBTYPE=BUSY:20100317T123000/20100317T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Gil Cohen (CS, Technion) about Theory Seminar: On the Degree of Symmetric Functions on the Boolean Cube at 2010-03-17 12:30:00 DESCRIPTION;LANGUAGE=en-US:In the 1997 paper "Polynomials with two values" by von zur Gathen and Roche (Combinatorica), the authors proved that th e degree of any non-constant symmetric function of the form $f:{0,1}^n \ri ghtarrow {0,1}$ is $n-o(n)$ (where the degree of a function is defined t o be the degree of the unique interpolation polynomial, of degree at mos t $n$, of the function). In their proof, the authors heavily used the fact that the function is Boolean. The authors therefore asked what can be s aid about the degree of functions of the form $f:{0,1}^n \rightarrow {0, 1,2,...,c}$.

While for $c=1$ the result above assures us all suc h functions have high degree, it is easy to see that for $c=n$ there exi st a function with degree 1. The authors proved a lower bound of $(n+1)/(c +1)$ for general $c$.

In this talk we show what we consider to be an interesting threshold phenomenon: For every $c
Joint work with Amir Shpi lka. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec441040008560 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100317T143000 DTEND;TZID="Asia/Jerusalem":20100317T163000 DTSTAMP;TZID="Asia/Jerusalem":20100317T143000 FREEBUSY;FBTYPE=BUSY:20100317T143000/20100317T163000 SUMMARY;LANGUAGE=en-US:The phd talk by Itay Maman about Code Structures in Java at 2010-03-17 14:30:00 DESCRIPTION;LANGUAGE=en-US:This research explores the identification of re curring constructs, Code Structures, which serve as the building blocks of object oriented programs and the exploitation of these in various area s including software research, development tools, and language design. Any in-depth research of these structures, be it an empirical stud y of existing programs or an analytical reasoning about software module s, must be based on some formal means for describing the elements of a program. To this end we present the Java Tools Language, JTL, which a rela tional query language that is specifically designed for querying the do main of Java elements. Using JTL we investigate existing Java programs, and present a comprehensive catalog of Micro Patterns capturing purposefu l programming idioms. An empirical study allows us to make precise clai ms regarding the prevalence of the patterns and the ability of these to characterize the software containing them. We also show how Code Str uctures can be used to enhance Java's type system with the ability to expr ess types whose conformance semantics is that of structural equivalence . This is realized by Whiteoak: A Java extension in which structural ty pes are defined via a strict subset of JTL's syntax. Thanks to these types Whiteoak can express a wider range of programs than Java programs with out compromising Java's static type safety. We also show how structural conformance can be used to support constructs such as virtual constructor s or anonymous types. Our last result is that of extending JTL suc h that it can produce output thereby allowing the JTL user to easily tr ansform or translate existing programs. This extension maintains JTL's pur ely logical nature. In particular production of output did not require the introduction of side effects into the language core. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012050 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100318T113000 DTEND;TZID="Asia/Jerusalem":20100318T133000 DTSTAMP;TZID="Asia/Jerusalem":20100318T113000 FREEBUSY;FBTYPE=BUSY:20100318T113000/20100318T133000 SUMMARY;LANGUAGE=en-US:The colloq talk by Isolde Adler SPECIAL LECTURE not e unusual hour about Model Checking Using Tree-width Methods: Old and New at 2010-03-18 11:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 601 Taub Bld. UID:111ec441040008620 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100325T143000 DTEND;TZID="Asia/Jerusalem":20100325T163000 DTSTAMP;TZID="Asia/Jerusalem":20100325T143000 FREEBUSY;FBTYPE=BUSY:20100325T143000/20100325T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Yuri Gurevich about The security tower of Babel at 2010-03-25 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008490 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100406T143000 DTEND;TZID="Asia/Jerusalem":20100406T163000 DTSTAMP;TZID="Asia/Jerusalem":20100406T143000 FREEBUSY;FBTYPE=BUSY:20100406T143000/20100406T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Dror Rawitz about Online Set Pac king and Competitive Scheduling of Multi-Part Tasks at 2010-04-06 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008600 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100407T123000 DTEND;TZID="Asia/Jerusalem":20100407T143000 DTSTAMP;TZID="Asia/Jerusalem":20100407T123000 FREEBUSY;FBTYPE=BUSY:20100407T123000/20100407T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Yotam Elor (CS, Technion ) about Theory Seminar: A Thermodynamic Approach to the Analysis of Multi- Robot Cooperative Localization under Independent Errors at 2010-04-07 12:3 0:00 DESCRIPTION;LANGUAGE=en-US:Abstract: We propose a new approach to the simu ltaneous cooperative localization of a group of robots capable of sensin g their own motion and the relative position of nearby robots. In the last decade, the use of distributed optimal Kalman filters (KF) to solve thi s problem have been studied extensively. In this paper, we propose to us e a sub-optimal Kalman filter (denoted by EA). EA requires significantly l ess computation and communication resources then KF. Furthermore, in som e cases, EA provides better localization.

In this paper EA is analyzed in a soft “thermodynamic” fashion i.e. relaxing assumptions a re used during the analysis. The goal is not to derive hard lower or upp er bounds but rather to characterize the robots expected behavior. In pa rticular, to predict the expected localization error. The predictions were validated using simulations. We believe that this kind of analysis can be beneficial in many other cases. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec441040008570 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100407T130000 DTEND;TZID="Asia/Jerusalem":20100407T150000 DTSTAMP;TZID="Asia/Jerusalem":20100407T130000 FREEBUSY;FBTYPE=BUSY:20100407T130000/20100407T150000 SUMMARY;LANGUAGE=en-US:The msc talk by Kira Radinsky about Temporal Concep t Learning (or how to predict the future?) at 2010-04-07 13:00:00 DESCRIPTION;LANGUAGE=en-US:Man has long desired to predict events that are likely to take place in the future. Such predictions can be beneficial f or planning, resource allocation and identification of risks and opportuni ties. Predicting events in politics, economics, society, etc. is a diffic ult task that is usually performed by human experts possessing extensive domain-specific and common-sense knowledge. In this research we aim to l earn how to predict in temporal environments with dynamically changing fe atures, labels and temporal correlations with other objects in the domain. We develop a general framework for temporal concept learning, as a gener alization of the traditional learning architecture, and present a family of algorithms for efficient concept learning in temporal domains. We pres ent experiments on several real-world prediction tasks, such as foreign ex change, weather, oil trading and hurricane prediction, that confirm the superior performance of our method compared to the previous state of the a rt methods. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400012080 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100412T183000 DTEND;TZID="Asia/Jerusalem":20100412T203000 DTSTAMP;TZID="Asia/Jerusalem":20100412T183000 FREEBUSY;FBTYPE=BUSY:20100412T183000/20100412T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Omer Boehm about Ha ifux, Haifa Linux Club: Genetic Algorithms at 2010-04-12 18:30:00 DESCRIPTION;LANGUAGE=en-US:TBA ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec441040008630 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100413T143000 DTEND;TZID="Asia/Jerusalem":20100413T163000 DTSTAMP;TZID="Asia/Jerusalem":20100413T143000 FREEBUSY;FBTYPE=BUSY:20100413T143000/20100413T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Daniel Marx about Movement Probl ems at 2010-04-13 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008660 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100421T143000 DTEND;TZID="Asia/Jerusalem":20100421T163000 DTSTAMP;TZID="Asia/Jerusalem":20100421T143000 FREEBUSY;FBTYPE=BUSY:20100421T143000/20100421T163000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Nir Atias (The Bla vatnik School of Computer Science, Tel Aviv University) about Bioinformati cs Forum: An Algorithmic Framework for Predicting Side-Effects of Drugs at 2010-04-21 14:30:00 DESCRIPTION;LANGUAGE=en-US:One of the critical stages in drug development is the identification of potential side effects for promising drug leads . Large scale clinical experiments aimed at discovering such side effect s are very costly and may miss subtle or rare side effects. To date, and to the best of our knowledge, no computational approach was suggested t o systematically tackle this challenge. In this work we report on a nove l approach to predict the side effects of a given drug. Starting from a query drug, a combination of canonical correlation analysis and network- based diffusion are applied to predict its side effects.

We e valuate our method by measuring its performance in cross validation using a comprehensive data set of 692 drugs and their known side effects deriv ed from package inserts. For 34% of the drugs the top scoring side effec t matches a known side effect of the drug. Remarkably, even on unseen da ta, our method is able to infer side effects that highly match existing knowledge. Our method thus represents a promising first step toward shor tcutting the process and reducing the cost of side effect elucidation. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008640 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100426T183000 DTEND;TZID="Asia/Jerusalem":20100426T203000 DTSTAMP;TZID="Asia/Jerusalem":20100426T183000 FREEBUSY;FBTYPE=BUSY:20100426T183000/20100426T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Orr Dunkelman (Weiz mann Institute) about Haifux, Haifa Linux Club: Making the Internet Access ible - workshop at 2010-04-26 18:30:00 DESCRIPTION;LANGUAGE=en-US:In this meeting we shall try and contact as man y webmasters of as many websites that do not support Linux-based browsing (as well as OpenSource browsing in other OSes). The aim is to offer a grea t deal of "demand" to various websites, thus proving the importance of sup porting wide range of platforms (and following the internet standards).

Before the meeting, we shall have a list of websites and the c ontact details of the webmasters (including by phone). If you have a candi date, please drop Orr a line (at orr(dot)dunkelman(at)gmail, and do not fo rget to add .com where relevant). ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec441040008680 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100428T113000 DTEND;TZID="Asia/Jerusalem":20100428T133000 DTSTAMP;TZID="Asia/Jerusalem":20100428T113000 FREEBUSY;FBTYPE=BUSY:20100428T113000/20100428T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Albert Cohen (Laboratoire J. -L. Lions, University Paris VI) about Pixel Club Seminar: Theory and Algor ithms for Anisotropic Triangulations with Applications to Image Representa tions at 2010-04-28 11:30:00 DESCRIPTION;LANGUAGE=en-US:We present the first results of an ongoing proj ect revolving around approximation by finite element functions on adaptive and anisotropic triangulations, with application to image processing.

We first recall the available theory for isotropic triangula tions which involves Besov-Sobolev spaces. For anisotropic triangulations, we present an analytic criterion that governs the rate of convergence in Lp norms for optimally adapted triangulations. We propose a greedy algorit hm which has the ability to generate triangulations that exhibit a locally optimal aspect ratio and prove that the optimal convergence rate is met b y the algorithm. We also present applications to image representation and compression. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008670 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100504T113000 DTEND;TZID="Asia/Jerusalem":20100504T133000 DTSTAMP;TZID="Asia/Jerusalem":20100504T113000 FREEBUSY;FBTYPE=BUSY:20100504T113000/20100504T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Arik Yavnai about Pixel Club Seminar:
נחילים – מעולם הטבע לעולם ההנדסה – מה ניתן ללמוד מהטבע at 2010-05-04 11:30:00 DESCRIPTION;LANGUAGE=en-US:חקירה והתבוננות בעולם הטב ע חוללה תובנות רבות בקרב החוקרים. נחילים בטבע, כמו גם בעלי חיים החיים בקבוצות שיתו פיות, מציגים יכולות מרשימות בהשגת יעדים קבוצתיים, גם אם יכולות הפרטים בקהילה מוג לות למדי.

התמודדות עם בעיות מורכב ות מדרבנת את החיפוש אחר פתרונות הכרוכים הפעלה של מערכים הנדסיים מרובי כלים – "נח ילים הנדסיים". מטרתם היא השגת יעדים משותפ ים תוך התמודדות עם סביבות לא מוכרות, אי-ו דאויות, דינמיקה של אירועים, והפתעות תרחי ש.

ההרצאה תציג תובנות מהטבע, את המ טיבציה לפיתוח מערכות הנדסיות מרובות כלי ם, וסוגיות שונות הכרוכות בתהליכי התכן וה ימוש של מערכות כאלו ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec441040008760 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100504T143000 DTEND;TZID="Asia/Jerusalem":20100504T163000 DTSTAMP;TZID="Asia/Jerusalem":20100504T143000 FREEBUSY;FBTYPE=BUSY:20100504T143000/20100504T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Lior Wolf about The Identificati on of Join Candidates in the Cairo Genizah: An interdisciplinary challenge in computer assisted text processing at 2010-05-04 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008700 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100505T143000 DTEND;TZID="Asia/Jerusalem":20100505T163000 DTSTAMP;TZID="Asia/Jerusalem":20100505T143000 FREEBUSY;FBTYPE=BUSY:20100505T143000/20100505T163000 SUMMARY;LANGUAGE=en-US:The phd talk by Tamir Levi about Bitonic Sorters Of Minimal Depth at 2010-05-05 14:30:00 DESCRIPTION;LANGUAGE=en-US:A Bitonic sorter is a comparator network that s orts every Bitonic input sequence. In our research, we studied the minim al depth of such networks. Building on previous works, we establish that the minimal depth of a Bitonic sorter of n keys is 2ceiling(log(n)) -flo or(log(n)). This result is constructive - that is, we present, for every n, a Bitonic sorter of that depth. In this talk I will also present oth er results from my PhD research. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012100 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100505T153000 DTEND;TZID="Asia/Jerusalem":20100505T173000 DTSTAMP;TZID="Asia/Jerusalem":20100505T153000 FREEBUSY;FBTYPE=BUSY:20100505T153000/20100505T173000 SUMMARY;LANGUAGE=en-US:The msc talk by Daniel Genkin about Radical Lexical ization of Mildly Context-Sensitive Languages at 2010-05-05 15:30:00 DESCRIPTION;LANGUAGE=en-US:A family of languages is called mildly context- sensitive if 1. it includes the family of all context-free languages; 2. it contains the languages * L= { a^n b^n c^n : n > 0 } - multiple ag reement, * L= { a^ m b^n c^m d^n : m,n > 0 } - crossed dependencies, * L= {ww : |w| > 0 } - reduplication ; 3. all its languages are se mi-linear; and 4. their membership problem is decidable in polynomial tim e. In our work we introduce a new model of computation called buffer a ugmented pregroup grammars that defines a family of mildly context-sensit ive languages. This model of computation is an extension of Lambek pregr oup grammars with a variable symbol- the buffer and is allowed to make an arbitrary substitution from the original pregroup to the variable. Thi s increases the pregroup grammar recognition power, but still retains the desired properties of mildly context-sensitive languages such as semi-l inearity and polynomial parsing. We establish a strict hierarchy within t he family of mildly context-sensitive languages defined by buffer augmen ted pregroup grammars. In this hierarchy, the hierarchy level of the fa mily language depends on the allowed number of occurrences of the variab le in word assignments. We also present an automaton counterpart of bu ffer augmented pregroup grammars, called buffer augmented pushdown automa ta and prove that a language is generated by a buffer augmented pregrou p grammar if and only if it is accepted by a buffer augmented pushdown automaton. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400011940 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100510T113000 DTEND;TZID="Asia/Jerusalem":20100510T133000 DTSTAMP;TZID="Asia/Jerusalem":20100510T113000 FREEBUSY;FBTYPE=BUSY:20100510T113000/20100510T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Anastasia Dubrovina (EE, Tec hnion) about Pixel Club Seminar: Non-Rigid Shape Correspondence by Matchin g Semi-Local Spectral Features and Global Geodesic Structures at 2010-0 5-10 11:30:00 DESCRIPTION;LANGUAGE=en-US:Non-rigid shape correspondence by matching semi -local spectral features and global geodesic structures Abstract: We prese nt an efficient computational method for finding correspondences between n on-rigid shapes. It utilizes both pointwise surface descriptors, and metri c structures defined on the shapes to perform the matching task, which is formulated as a quadratic minimization problem. The suggested surface desc riptors are based on eigendecomposition of the Laplace-Beltrami operator. It is also observed that when using the above descriptors, multiple possib le correspondences may exist between two shapes that are intrinsically sym metric. It follows from the fact that the eigenfunctions of the Laplace-Be ltrami operator defined on a shape are symmetric with respect to its intri nsic symmetries. Consequently, we present an algorithm to construct differ ent descriptors for each possible correspondence. When used in a proper mi nimization problem, those descriptors allow us to explode all possible cor respondences between two given shapes. The lecture will take place on Mond ay, May 10, 2010, 11.30 at EE Meyer Building 1061 Electrical Engineering D epartment. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec441040008820 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100510T183000 DTEND;TZID="Asia/Jerusalem":20100510T203000 DTSTAMP;TZID="Asia/Jerusalem":20100510T183000 FREEBUSY;FBTYPE=BUSY:20100510T183000/20100510T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Roy Migdal (CS, Tec hnion) about Haifux, Haifa Linux Club: Compromising Electromagnetic Emanat ions of Wired and Wireless Keyboards at 2010-05-10 18:30:00 DESCRIPTION;LANGUAGE=en-US:Roy Migdal will present the work of Martin Vua gnoux and Sylvain Pasini, which won the best paper award at 18th USENIX Se curity Symposium (Usenix Security '09), Montreal, Canada, August 10-14, 20 09.

Computer keyboards are often used to transmit confiden tial data such as passwords. Since they contain electronic components, key boards eventually emit electromagnetic waves. These emanations could revea l sensitive information such as keystrokes. The technique generally used t o detect compromising emanations is based on a wide-band receiver, tuned o n a specific frequency. However, this method may not be optimal since a si gnificant amount of information is lost during the signal acquisition. Our approach is to acquire the raw signal directly from the antenna and to pr ocess the entire captured electromagnetic spectrum. Thanks to this method, we detected four different kinds of compromising electromagnetic emanatio ns generated by wired and wireless keyboards. These emissions lead to a fu ll or a partial recovery of the keystrokes. We implemented these side-chan nel attacks and our best practical attack fully recovered 95\% of the keys trokes of a PS/2 keyboard at a distance up to 20 meters, even through wall s. We tested 12 different keyboard models bought between 2001 and 2008 (PS /2, USB, wireless and laptop). They are all vulnerable to at least one of the four attacks. We conclude that most of modern computer keyboards gener ate compromising emanations (mainly because of the manufacturer cost press ures in the design). Hence, they are not safe to transmit confidential inf ormation. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec441040008880 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100511T143000 DTEND;TZID="Asia/Jerusalem":20100511T163000 DTSTAMP;TZID="Asia/Jerusalem":20100511T143000 FREEBUSY;FBTYPE=BUSY:20100511T143000/20100511T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Danny Raz about Oblivious and Co st Aware Distributed Load Sharing in the Cloud at 2010-05-11 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008840 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100524T183000 DTEND;TZID="Asia/Jerusalem":20100524T203000 DTSTAMP;TZID="Asia/Jerusalem":20100524T183000 FREEBUSY;FBTYPE=BUSY:20100524T183000/20100524T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Dotan Cohen (Techni on) about Haifux, Haifa Linux Club: KDE 4 - The Good, The Bad, and The Bro ken at 2010-05-24 18:30:00 DESCRIPTION;LANGUAGE=en-US:KDE was developed to give Unix applications a c ommon look and feel. From 1996 to 2007 the desktop environment saw many im provements and changes. In January 2008 a rewrite of the entire desktop en vironment and many of it's core applications was performed. The rewrite, c alled KDE 4, led to a dramatic change in KDE's stance. Once noted for its configurablity, KDE 4 had significantly fewer options than previous KDE ve rsions. Core applications and the desktop environment itself were missing critical features. Crashes were common. One year later KDE 4.2 was introdu ced, promising to address all these issues. Additionally, several innovati ve technologies were introduced then. In this talk we will examine KDE 4.4 and concentrate on one innovative feature (the Good), one questionable de sign decision (the Bad), and one critical bug (the Broken). ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec441040008960 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100525T113000 DTEND;TZID="Asia/Jerusalem":20100525T133000 DTSTAMP;TZID="Asia/Jerusalem":20100525T113000 FREEBUSY;FBTYPE=BUSY:20100525T113000/20100525T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Evgeny Gershikov (Technion) about Pixel Club Seminar: On the Role of Color Information in Image Proces sing and Visual Communication at 2010-05-25 11:30:00 DESCRIPTION;LANGUAGE=en-US:Color information plays a major role in image p rocessing and visual communication although presently most algorithms and tools are developed mainly for monochromatic images. Usually, the processi ng of color images is performed either in the RGB color space or in anothe r color space chosen rather arbitrarily, such as YUV or YIQ. In this work we propose new frameworks for color image processing and coding based on a n optimized approach. These frameworks along with rate-distortion analysis optimize the stages of coding and color processing for visual communicati on. The Mean Square Error (MSE( and the visual-oriented Weighted Mean Squa re Error (WMSE) are used to account for both quantitative and subjective v isual fidelity. We exploit the high inter-color correlations of the RGB pr imaries to introduce a correlation-based coding approach rather than the o rdinary decorrelation-based method, which is currently used in most algori thms. The two approaches are further generalized to provide a unified corr elation/decorrelation-based framework.

The new color processi ng approach can be helpful in several fields of image processing, includin g image compression and image demosaicing. We show that for compression, t he new framework outperforms presently available compression algorithms, i ncluding well established ones, such as JPEG 2000, while having comparable complexity. For demosaicing, instead of using the ordinary RGB color spac e, the correlation of primary colors can be better exploited in an optimiz ed color space, resulting in demosaicing that is superior to presently ava ilable methods. Additional aspects of the new frameworks will be presented and discussed. Our conclusion is that by optimizing the use of color info rmation, major operations in image processing and visual communication cou ld be improved quantitatively and visually. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec441040008900 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100525T143000 DTEND;TZID="Asia/Jerusalem":20100525T163000 DTSTAMP;TZID="Asia/Jerusalem":20100525T143000 FREEBUSY;FBTYPE=BUSY:20100525T143000/20100525T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Michael Bronstein about Computat ional Metric Geometry: A New Tool in Image Sciences at 2010-05-25 14:3 0:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008950 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100526T140000 DTEND;TZID="Asia/Jerusalem":20100526T160000 DTSTAMP;TZID="Asia/Jerusalem":20100526T140000 FREEBUSY;FBTYPE=BUSY:20100526T140000/20100526T160000 SUMMARY;LANGUAGE=en-US:The phd talk by Arik Friedman about Privacy Preserv ing Data Mining at 2010-05-26 14:00:00 DESCRIPTION;LANGUAGE=en-US:In recent years the data mining community has f aced a new challenge. Having shown how effective its tools are in revealin g the knowledge locked within huge databases, it is now required to develo p methods that restrain the power of these tools to protect the privacy of individuals. This research focuses on the problem of guaranteeing priv acy of data mining output. In the talk I will present some of the recent r esearch results. We consider the problem of data mining with formal privac y guarantees, given a data access interface based on the differential priv acy framework. Differential privacy requires that computations be insensit ive to changes in any particular individual's record, thereby restricting data leaks through the results. The privacy preserving interface ensures u nconditionally safe access to the data and does not require from the data miner any expertise in privacy. However, a naive utilization of the interf ace to construct privacy preserving data mining algorithms could lead to i nferior data mining results. We address this problem by considering the privacy and the algorithmic requirements simultaneously, focusing on decis ion tree induction as a sample application. The privacy mechanism has a profound effect on the performance of the methods chosen by the data miner . We demonstrate that this choice could make the difference between an acc urate classifier and a completely useless one. Moreover, an improved algor ithm can achieve the same level of accuracy and privacy as the naive imple mentation but with an order of magnitude fewer learning samples. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400011920 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100526T163000 DTEND;TZID="Asia/Jerusalem":20100526T183000 DTSTAMP;TZID="Asia/Jerusalem":20100526T163000 FREEBUSY;FBTYPE=BUSY:20100526T163000/20100526T183000 SUMMARY;LANGUAGE=en-US:The Mathematics Club talk by Ron Kimmel (CS, Techni on) about Mathematics Club Seminar: Metric Geometry in Action at 2010-05-2 6 16:30:00 DESCRIPTION;LANGUAGE=en-US:More detail in seminar poster
פרטים נוספים ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Amado 232 UID:111ec441040008990 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100531T113000 DTEND;TZID="Asia/Jerusalem":20100531T133000 DTSTAMP;TZID="Asia/Jerusalem":20100531T113000 FREEBUSY;FBTYPE=BUSY:20100531T113000/20100531T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Radu Horaud (INRIA Rhone-Alp es, Grenoble, France) about Pixel Club Seminar: 3D Shape Representation Us ing Diffusion Embeddings at 2010-05-31 11:30:00 DESCRIPTION;LANGUAGE=en-US:We will address the problem of representing sha pes using eigenvalues and eigenvectors of the discrete diffusion operator. A discrete shape, such as a mesh or a point cloud, can be viewed as an un directed weighted graph, hence one can use spectral graph theory to both embed and analyse shapes. We propose two graph diffusion operators that a re built based on two widely used graph Laplacians: The combinatorial Lapl acian and the normalized Laplacian. We recapitulate the basic spectral pro perties of these matrices and we thoroughly motivate the choice of the com binatorial Laplacian. The embedded shape is therefore treated as a distrib ution in the corresponding embedded (or feature) space. We propose several possible normalizations and we characterize some of their statistical pro perties. We will illustrate the usefulness of this shape representations f or the task of sparse shape matching. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec441040008970 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100601T113000 DTEND;TZID="Asia/Jerusalem":20100601T133000 DTSTAMP;TZID="Asia/Jerusalem":20100601T113000 FREEBUSY;FBTYPE=BUSY:20100601T113000/20100601T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Radu Horaud (INRIA Rhone-Alp es, Grenoble, France) about Pixel Club Seminar: Audio - Visual Fusion at 2 010-06-01 11:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec441040009550 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100601T143000 DTEND;TZID="Asia/Jerusalem":20100601T163000 DTSTAMP;TZID="Asia/Jerusalem":20100601T143000 FREEBUSY;FBTYPE=BUSY:20100601T143000/20100601T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Moshe Y. Vardi about From Philos ophical to Industrial Logics at 2010-06-01 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008800 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100601T143000 DTEND;TZID="Asia/Jerusalem":20100601T163000 DTSTAMP;TZID="Asia/Jerusalem":20100601T143000 FREEBUSY;FBTYPE=BUSY:20100601T143000/20100601T163000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Richard Tsai (Mathematics, U niversity of Texas at Austin) about Pixel Club Seminar: Inverse Source Pro blems in Domains with Holes at 2010-06-01 14:30:00 DESCRIPTION;LANGUAGE=en-US:We consider the inverse problem of discovering the location of point sources from very sparse point measurements in a bou nded domain that contains impenetrable (and possibly unknown) obstacles.

We present our adaptive algorithm for determining the measurem ent locations, and ultimately, the source locations. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec441040008980 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100602T133030 DTEND;TZID="Asia/Jerusalem":20100602T153030 DTSTAMP;TZID="Asia/Jerusalem":20100602T133030 FREEBUSY;FBTYPE=BUSY:20100602T133030/20100602T153030 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Prof. David Horn ( School of Physics & Astronomy, Tel Aviv University) about Bioinformatics F orum: Deriving Enzymatic and Taxonomic Signatures of Metagenomes from S hort Read Data at 2010-06-02 13:30:30 DESCRIPTION;LANGUAGE=en-US:We propose a method for deriving enzymatic sign atures from short read (SR) metagenomic data of unknown species. The SR data are converted to six pseudo-peptide candidates. We search for occur rences of Specific Peptides (SPs) on the latter. SPs are peptides that a re indicative of enzymatic function as defined by the Enzyme Commission (EC) nomenclature. Counting their hits, we associate short reads with sp ecific EC categories. The putative peptide counts can then be converted to estimates of numbers of enzymes associated with the given EC categori es in the studied metagenome, thus defining its enzymatic spectrum witho ut the need to perform genomic assemblies of short reads. The method is developed and tested on 22 bacteria for which there exist good EC annota tions in NCBI. Enzymatic signatures are derived for 3 metagenomes, and t heir functional profiles are explored. We extend the SP methodology to t axon-specific SPs (TSPs), allowing us to estimate also taxonomic features of metagenomic data from short-reads. Using recent Swiss-Prot data we ob tain TSPs for different phyla of bacteria, and different classes of prot eobacteria. These allow us to analyze leading taxa content of 4 differen t metagenomic datasets.

Joint work with Uri Weingart, Erez Pe rsi and Uri Gophna. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec441040008920 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100603T130000 DTEND;TZID="Asia/Jerusalem":20100603T150000 DTSTAMP;TZID="Asia/Jerusalem":20100603T130000 FREEBUSY;FBTYPE=BUSY:20100603T130000/20100603T150000 SUMMARY;LANGUAGE=en-US:The cggc talk by Joab Winkler (Sheffield University , Sheffield, England about CGGC Seminar: A Polynomial Root Solver that pr eserves The Multiplicities of the Roots at 2010-06-03 13:00:00 DESCRIPTION;LANGUAGE=en-US:TBA ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec441040008890 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100607T183000 DTEND;TZID="Asia/Jerusalem":20100607T203000 DTSTAMP;TZID="Asia/Jerusalem":20100607T183000 FREEBUSY;FBTYPE=BUSY:20100607T183000/20100607T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Orna Agmon Ben-Yehu da (CS, Technion) about Haifux, Haifa Linux Club: Gupta et al.'s paper (an d work) on Difference Engine. at 2010-06-07 18:30:00 DESCRIPTION;LANGUAGE=en-US:Difference Engine is an extension to Xen, which is an open-source hypervisor. The paper, titled "Difference Engine: Harne ssing Memory Redundancy in Virtual Machines", by Diwaker Gupta, Sangmin Le e,Michael Vrable, Stefan Savage, Alex C. Snoeren, George Varghese, Geoffre y M. Voelker, and Amin Vahdat, won a best paper award in OSDI'08. I will n ot present the conference talk (which is just a taste, and for which slide s are unavailable) but the full paper, including my comments on it. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec441040009580 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100608T113000 DTEND;TZID="Asia/Jerusalem":20100608T133000 DTSTAMP;TZID="Asia/Jerusalem":20100608T113000 FREEBUSY;FBTYPE=BUSY:20100608T113000/20100608T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Emil Saucan (Mathematics, Te chnion) about Pixel Club Seminar: Nash, Burago & Zalgaller and Image Proce ssing at 2010-06-08 11:30:00 DESCRIPTION;LANGUAGE=en-US:The celebrated Nash Embedding Theorem is extens ively employed in recent years in various theoretical, as well as practica l, aspects of Computer Graphics and Image Processing. However, in practice , this kind of application encounters certain obstructions, rendering the use of the Embedding Theorem as somewhat problematic. We explore therefore its practicability in Vision, Graphics and other Imaging sciences. As a solution for some of the problems mentioned above, the use of a PL versio n of the Theorem, due to Burago and Zalgaller, regarding the existence of isometric embeddings of polyhedral surfaces in R3, has been suggested. We briefly examine the proof of this last result and we show that their proof does not extended directly to higher dimensions.

In addit ion, a number of possible adaptations, relaxations and somewhat different approaches to the Embedding Theorem, rendering it more practical for imagi ng, are also considered. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec441040008910 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100608T143000 DTEND;TZID="Asia/Jerusalem":20100608T163000 DTSTAMP;TZID="Asia/Jerusalem":20100608T143000 FREEBUSY;FBTYPE=BUSY:20100608T143000/20100608T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Yishay Mansour - Annual S. Even Distinguished Colloquium about Regret Minimization and Job Scheduling at 2010-06-08 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008720 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100609T110000 DTEND;TZID="Asia/Jerusalem":20100609T130000 DTSTAMP;TZID="Asia/Jerusalem":20100609T110000 FREEBUSY;FBTYPE=BUSY:20100609T110000/20100609T130000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Asa Ben-Hur (Depar tment of Computer Science, Colorado State University) about Bioinformatics Forum: Accurate Protein Function Prediction with the GOstruct Method at 2 010-06-09 11:00:00 DESCRIPTION;LANGUAGE=en-US:Protein function prediction is an active area o f research in bioinformatics. And yet, transfer of annotation on the bas is of sequence or structural similarity remains widely used in practice. Most of the machine learning methods applied to this problem reduce it to a collection of binary classification problems: whether a protein per forms a particular function, sometimes with a post-processing step to co mbine the binary outputs. I will present the GOstruct method that direct ly predicts the Gene Ontology terms associated with a protein by modelin g the hierarchical structure of the Gene Ontology in the framework of ke rnel methods for structured-output spaces. Empirical results show improved performance over several algorithms that were recently benchmarked on the Mousefunc dataset.

Host: Tomer Shlomi ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec441040009640 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100609T113000 DTEND;TZID="Asia/Jerusalem":20100609T133000 DTSTAMP;TZID="Asia/Jerusalem":20100609T113000 FREEBUSY;FBTYPE=BUSY:20100609T113000/20100609T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Eitan Grinspun (Computer Sci ence, Columbia University) about Pixel Club Seminar: "Good" Computation f or the Motion of Colliding Objects at 2010-06-09 11:30:00 DESCRIPTION;LANGUAGE=en-US:We would like to compute the motion of flexible materials colliding against each other in complex ways (e.g., sheets of f abric being knotted, plastic and metal containers crushed in a trash compa ctor). Such computations are needed in special effects, engineering design , medical simulation, and any other domain that requires consideration of materials interacting against each other. What is a "good" way to carry ou t such computations?

What sets our approach apart from previo us attempts is our focus on three guarantees: that simulations of well-pos ed problems (a) have no interpenetrations, (b) obey causality, momentum- a nd energy-conservation laws, and (c) complete in finite time. While these three requirements may seem natural to ask of a computation, we will show that previous methods can be grouped into three families, each family sacr ificing one of the three guarantees for the sake of the other two. Various 2D and 3D examples will illustrate why a single method offering all three guarantees can serve as a powerful simulation tool.

This tal k will be given at a tutorial level, without assuming prior knowledge of m otion computation or collision response algorithms.

*This is joint work with David Harmon, Etienne Vouga, Breannan Smith, and Rasmus Ta mstorf ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec441040009630 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100609T133000 DTEND;TZID="Asia/Jerusalem":20100609T153000 DTSTAMP;TZID="Asia/Jerusalem":20100609T133000 FREEBUSY;FBTYPE=BUSY:20100609T133000/20100609T153000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Dr. Sergey Ermakov (Sackler Faculty of Medicine, Tel Aviv University) about Bioinformatics F orum: Useful Tools in Finding Genetic Determinants of Complex Traits: Qu antitative Genetic Analysis Software and Bioinformatics Resources for Mark er Prioritization. at 2010-06-09 13:30:00 DESCRIPTION;LANGUAGE=en-US:Substantial familial aggregation and heritabili ty is observed for a wide range of phenotypes in general human populatio n. Much effort has been made to decipher which specific genetic factors contribute to normal interindividual variability and development of path ologic conditions. The lecture will focus on the workflow of a typical f amily-based genetic research, illustrated by the association study of RU NX2 polymorphisms and hand bone length and BMD, relevant to osteoporosis . The Program Package MAN for pedigree analysis of quantitative traits a nd genetic markers in human pedigrees will be presented as well as a num ber of bioinformatic tools useful within the framework of genetic associ ation studies, i.e. overview of the marker properties and tagging SNP se lection (HaploView), marker prioritization (FastSNP, Pupasuite, SNPinfo, RegulatoryGenomics.org, SCAN, mRNA by SNP Browser), analysis of the study results in the genomic context (UCSC genome browser). ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec441040009560 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100609T133030 DTEND;TZID="Asia/Jerusalem":20100609T153030 DTSTAMP;TZID="Asia/Jerusalem":20100609T133030 FREEBUSY;FBTYPE=BUSY:20100609T133030/20100609T153030 SUMMARY;LANGUAGE=en-US:The cryptoday2010 talk by Workshop in Cryptology ab out CRYPTODAY2010 at 2010-06-09 13:30:30 DESCRIPTION;LANGUAGE=en-US:The 2010 Workshop in Cryptology will be held on Wednesday, June 9 2010, between 9:00-16:10, in Auditorium 1, CS Taub B uilding, Technion. You are all invited.

More details and program< /a> ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 1 UID:111ec441040008940 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100615T143000 DTEND;TZID="Asia/Jerusalem":20100615T163000 DTSTAMP;TZID="Asia/Jerusalem":20100615T143000 FREEBUSY;FBTYPE=BUSY:20100615T143000/20100615T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Distinguished professor Leonard Kleinrock about My Life and My Work at 2010-06-15 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040008780 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100616T110000 DTEND;TZID="Asia/Jerusalem":20100616T130000 DTSTAMP;TZID="Asia/Jerusalem":20100616T110000 FREEBUSY;FBTYPE=BUSY:20100616T110000/20100616T130000 SUMMARY;LANGUAGE=en-US:The msc talk by Amit Bermano about Online reconstru ction of 3D objects from arbitrary cross-section data at 2010-06-16 11:00: 00 DESCRIPTION;LANGUAGE=en-US:The reconstruction of a surface from a set of p lanar cross-sections such that the surface interpolates, or approximates, the input has been thoroughly studied in the past decades. This problem a rises mainly in the fields of medical imaging (MRI, CT, ultrasound etc.) a nd geographical information systems (for terrain reconstruction). The inp ut is assumed to have been segmented in a preprocessing step, to create a set of closed two-dimensional contours, separating the ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400011870 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100616T143000 DTEND;TZID="Asia/Jerusalem":20100616T163000 DTSTAMP;TZID="Asia/Jerusalem":20100616T143000 FREEBUSY;FBTYPE=BUSY:20100616T143000/20100616T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Distinguished professor Leonard Kleinrock about A Brief History of the Internet at 2010-06-16 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 861 UID:111ec441040008790 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100620T110000 DTEND;TZID="Asia/Jerusalem":20100620T130000 DTSTAMP;TZID="Asia/Jerusalem":20100620T110000 FREEBUSY;FBTYPE=BUSY:20100620T110000/20100620T130000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Alfred Inselberg ( Mathematics, Tel Aviv University) about CGGC Seminar: Visualizing R^N – Representing Surfaces by their Normal Vectors at 2010-06-20 11:00:00 DESCRIPTION;LANGUAGE=en-US:With parallel coordinates the perceptual barrie r imposed by our 3-dimensional habitation is breached enabling the vi sualization of multidimensional problems. By learning to recognize patte rns a powerful knowledge discovery process evolved. It leads to a deeper geometrical insight: the recognition of M-dimensional objects recurs ively from their (M-1)-dimensional subsets. A hyperplane in N-dimensions is represented by (N -1) indexed points. Points representing lines ha ve two indices, those representing planes three indices and so on. In tu rn, this yields powerful geometrical algorithms (e.g. for intersections, containment, proximities) and applications including classification. A smooth surface in 3-D is the envelope of its tangent planes each repre sented by 2 planar points. As a result it is represented by two planar r egions, and a hypersurface in N-dimensions by (N-1) regions. This is equ ivalent to representing a surface by its normal vectors. Developable sur faces are represented by curves revealing the surface characteristics. C onvex surfaces in any dimension are recognized by hyperbola-like regions. Non-orientable surfaces yield stunning patterns unlocking new geometrica l insights. Non-convexities like folds, bumps, concavities are not hidde n. The patterns persist in the presence of errors and that’s good news for applications opening the way for the exploration of massive dataset s. Applications of parallel coordinates include collision avoidance and conflict resolution algorithms for air traffic control (3 USA patents ), computer vision (USA patent), data mining (USA patent), decision supp ort and process control. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 401 UID:111ec441040009660 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100620T113000 DTEND;TZID="Asia/Jerusalem":20100620T133000 DTSTAMP;TZID="Asia/Jerusalem":20100620T113000 FREEBUSY;FBTYPE=BUSY:20100620T113000/20100620T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Vladimir Lumelsky (NASA, Uni versity of Maryland) about Pixel Club Seminar: Whole-Body Robot Sensing Is Prerequisite for Human-Robot Interaction and Teams at 2010-06-20 11:30:0 0 DESCRIPTION;LANGUAGE=en-US:While control via sensing in robotics and in hu man-robot interaction is important for both robot autonomous vehicles and arm manipulators, it is more prescient and less understood in robot arm manipulation tasks. Analysis of such systems suggests the following:

* Mutual safety of humans vis-a-vie robots vis-a-vie other objects in an unstructured environment dictates a need for massive covera ge of the whole robot body with sensors.
* Achieving robot motion pl anning skills sufficient for an unstructured setting dictates the same, wh ole-body sensing.
* Theory shows that whole-body robot sensing gives rise to interesting and efficient motion planning algorithms that compete with human skills.
* Whole-body sensing hardware and its control pr esent a variety of challenges to today’s electronics, material science, and embedded microprocessor control.
* Humans have cognitive d ifficulties controlling robots in real time – causing severe constraints for human-robot teams, such as a robot helper for the elderly or a robot assistant to an astronaut.
* There is a happy circumstance: in so me motion planning tasks humans can think better than robots, while in som e others robots can think better than humans. The latter is true only if t he robot is given enough sensing – and it points to a natural synergy of human-robot teams.

Satisfactory resolution of the issues involved promises to deliver robot systems that are, on the one hand, caut ious and friendly to their hosts and environment (think of the Azimov’s laws), and on the other, allow unsupervised automation in unpredictable se ttings. In this talk we will expand on the bullets above, trying to shape them in terms of the issues in the disciplines involved:
* Algori thms for motion planning – kinematics, computational geometry & topology issues
* Cognitive science – studying (poorly-understood) huma n motion planning skills
* Electronics – large-area sensing arr ays, sensitive skin design, tactile vs proximity sensing
* Materi al science – bendable and stretchable sensitive skins
* Contro l - tradeoffs between sensor density, array sizes, sensitivity of motion p lanning ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040009680 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100620T123000 DTEND;TZID="Asia/Jerusalem":20100620T143000 DTSTAMP;TZID="Asia/Jerusalem":20100620T123000 FREEBUSY;FBTYPE=BUSY:20100620T123000/20100620T143000 SUMMARY;LANGUAGE=en-US:The phd talk by Emilia Katz about Analysis and Dete ction of Interactions Among Aspects at 2010-06-20 12:30:00 DESCRIPTION;LANGUAGE=en-US:Aspect-oriented programming is becoming a commo n approach to extend object systems with modules that cross-cut the usu al class hierarchy. Aspects encapsulate treatment of concerns that othe rwise would be scattered within an underlying application, and tangled with code treating other concerns. Often, insertion of several aspect s into one system is desired and in that case the problem of interferenc e among the aspects might arise, even if each aspect individually woven is correct relative to its specification. In this type of interference, one aspect can prevent another from having the required effect on a wove n system. Such interference is defined and an incremental proof strategy based on off-line model checking pairs of aspects for a generic model expressing the specifications is presented. When an aspect is added to a library of noninterfering aspects, only its interaction with each of the aspects from the library needs to be checked. Such checks for each p air of aspects are proven sufficient to detect interference or establish interference freedom for any collection of aspects in a library. Sev eral issues are treated that extend the applicability of proof methods f or aspects. One extension is for treatment of interactions of aspects of all the categories, where the division of aspects to categories is d one according to the possible influence of the aspects on the base syst em computations. An extended specification for aspects, and a new verif ication method based on model checking are presented. They are used to e stablish the correctness of strongly-invasive aspects independently of any particular base program to which they may be woven. Additional e xtension is needed for the case when multiple aspects can share a join-poi nt. In this case they may, but do not have to, semantically interfere, and a specification refinement might be necessary to enable modular ver ification and interference detection among aspects even in the presence of shared join-points. An in depth analysis of aspect semantics and mu tual influence of aspects at a shared join-point is presented, in order to enable distinguishing between potential and actual interference among aspects at shared join-points. An interactive semi-automatic procedure for specification refinement is described, that will help users define the intended aspect behavior more precisely. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400010860 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100621T183000 DTEND;TZID="Asia/Jerusalem":20100621T203000 DTSTAMP;TZID="Asia/Jerusalem":20100621T183000 FREEBUSY;FBTYPE=BUSY:20100621T183000/20100621T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Asaf Bartov (Ben-Ye huda Project, Founder of Hebrew Literatuare Computing Society) about Haifu x, Haifa Linux Club: Technological Challanges in Ben Yehuda Project at 201 0-06-21 18:30:00 DESCRIPTION;LANGUAGE=en-US:פרוייקט בן-יהודה הוא מיזם התנדבותי אשר בונה ומתחזק מאגר חופשי של י ירה עברית שאינה מוגנת בזכויות יוצרים. ה הדורות האלקטרוניות של היצירות מוגשות לצ יבור בחינם, ללא פרסומות, והן נחלת הכלל (publ ic domain), וזמינות לכל שימוש, לרבות שימוש מסח רי.

על אף המראה הבסיסי והמיושן ש האתר הנוכחי, יש עבודה רבה מאחורי הקלעים, והתמודדות עם כמה אתגרים, טכנולוגיים ואח ים, שעל מקצתם אדבר.

סקירת תוכן הרצאה:

- היכרות עם פרויקט בן-יהו ה ואופן פעולתו - סקירה טכנית של האתר הנו כחי -- יתרונות ומגבלות - אתגרים טכנולוגי ם: - ייצוג טקסט על רבדיו - חיפוש - יי צוג קטלוג מתקדם - על בחירת כלי הפיתוח לג סה הבאה של האתר - הצגת מערכת EbyDict להפקת מ הדורה אלקטרונית של מילון אליעזר בן-יהודה . ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400010870 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100622T113000 DTEND;TZID="Asia/Jerusalem":20100622T133000 DTSTAMP;TZID="Asia/Jerusalem":20100622T113000 FREEBUSY;FBTYPE=BUSY:20100622T113000/20100622T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Rudi Primorac (EE, Technion) about Pixel Club Seminar: Color Video Coding using Spatio-Temporal Correl ation of Primary Colors at 2010-06-22 11:30:00 DESCRIPTION;LANGUAGE=en-US:Natural videos, given in the RGB color space, a re characterized by high correlations between the primary colors (R, G and B). To reduce this redundancy, most video compression algorithms, like MP EG, decorrelate the color components by transforming the input video from the highly correlated RGB components into less correlated space such as YI Q or YCbCr.

In this work, we propose a different approach, wh ich utilizes the natural high inter-color correlation. We have developed a lgorithms for video coding that benefit from the correlation of the primar y colors in both the spatial and the frequency (DCT) domains. Video stream is subdivided into 3D blocks of variable size according to the contents a long the spatial and the temporal axes. In each block, one of the colors i s chosen as a 'base color', using a minimum reconstruction error criterion . While the 'base color' is coded using conventional tools, the two remain ing components are encoded as approximation polynomials of the base color.

Experimental results show that the proposed approach outperf orms presently available methods. Our conclusion is that the high correlat ion between primary RGB colors could be helpful for video coding and that the new spatio-temporal approach to video compression is more efficient th an conventional decorrelation-based techniques.

*MSc final pa per under the supervision of Dr. Moshe Porat ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec441040009690 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100622T143000 DTEND;TZID="Asia/Jerusalem":20100622T163000 DTSTAMP;TZID="Asia/Jerusalem":20100622T143000 FREEBUSY;FBTYPE=BUSY:20100622T143000/20100622T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Jure Leskovec about Large networ ks, Communities and Kronecker Products at 2010-06-22 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040009650 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100623T113000 DTEND;TZID="Asia/Jerusalem":20100623T133000 DTSTAMP;TZID="Asia/Jerusalem":20100623T113000 FREEBUSY;FBTYPE=BUSY:20100623T113000/20100623T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Polina Golland (MIT/CSAIL) a bout Pixel Club Seminar: Non-parametric Atlas-Based Segmentation of Highly Variable Anatomy at 2010-06-23 11:30:00 DESCRIPTION;LANGUAGE=en-US:We propose a non-parametric probabilistic model for the automatic segmentation of medical images. The resulting inferen ce algorithms register individual training images to the new image, tran sfer the segmentation labels and fuse them to obtain the final segmentat ion of the test subject. Our generative model yields previously proposed label fusion algorithms as special cases, but also leads to a new var iant that aggregates evidence locally in determining the segmentation la bels. We demonstrate the advantages of our approach in two clinical appl ication: segmentation of neuroanatomical structures and segmentation of the left heart atrium whose shape varies significantly across the popula tion. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400010910 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100624T113000 DTEND;TZID="Asia/Jerusalem":20100624T133000 DTSTAMP;TZID="Asia/Jerusalem":20100624T113000 FREEBUSY;FBTYPE=BUSY:20100624T113000/20100624T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Marcello Pelillo (University of Venice, Italy ) about Pixel Club Seminar: What is a Cluster? Perspec tives from Game Theory at 2010-06-24 11:30:00 DESCRIPTION;LANGUAGE=en-US:Contrary to the vast majority of approaches to clustering, which view the problem as one of partitioning a set of obse rvations into coherent classes, thereby obtaining the clusters as a by- product of the partitioning process, we propose to reverse the terms of the problem and attempt instead to derive a rigorous formulation of t he very notion of a cluster. In our endeavor to provide an answer to th is question, we found that game theory offers a very elegant and genera l perspective that serves well our purposes.

Accordingly, we formulate the clustering problem as a non-cooperative "clustering ga me". Within this context, the notion of a cluster turns out to be equiv alent to a classical equilibrium concept from (evolutionary) game theor y. Applications to computer vision problems and generalizations of the proposed idea will be discussed. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400010920 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100624T143000 DTEND;TZID="Asia/Jerusalem":20100624T163000 DTSTAMP;TZID="Asia/Jerusalem":20100624T143000 FREEBUSY;FBTYPE=BUSY:20100624T143000/20100624T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Distinguished Professor Izhak Ru bin about Cross Layer Networking and Management for Mobile Wireless Networ ks at 2010-06-24 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400010880 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100629T113000 DTEND;TZID="Asia/Jerusalem":20100629T133000 DTSTAMP;TZID="Asia/Jerusalem":20100629T113000 FREEBUSY;FBTYPE=BUSY:20100629T113000/20100629T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Tamar Back (EE, Technion) ab out Pixel Club Seminar: Incorporating Temporal Context in Bag-of-Words Mod els at 2010-06-29 11:30:00 DESCRIPTION;LANGUAGE=en-US:The Bag-of-Words (BoW) model is often used for recognition of objects, scenes, actions and more. It achieves impressive results in many diff erent areas, although it discards the spatial and temporal order of codewords in a labeled signal. This work is defi ning a new model: Contextual Sequence of Words (CSoW) which incorporates temp oral order in a BoW model for video representation, and tests it on acti on recognition tasks. The temporal context is incorporated in three scal es: global, medium and fine scale context. We show that using CSoW inst ead of BoW on the same setups achieves signifi cant improvements in action recognition rates, for 4 di fferent setups.

*MSc thesis un der the supervision of Lihi Zelnik Manor. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400010900 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100629T143000 DTEND;TZID="Asia/Jerusalem":20100629T163000 DTSTAMP;TZID="Asia/Jerusalem":20100629T143000 FREEBUSY;FBTYPE=BUSY:20100629T143000/20100629T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Hanan Samet about LOCATION, LOCA TION, LOCATION: CASTING A WIDE NET ON GEOGRAPHIC INFORMATION SYSTEMS (GIS) at 2010-06-29 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040009670 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100630T110000 DTEND;TZID="Asia/Jerusalem":20100630T130000 DTSTAMP;TZID="Asia/Jerusalem":20100630T110000 FREEBUSY;FBTYPE=BUSY:20100630T110000/20100630T130000 SUMMARY;LANGUAGE=en-US:The phd talk by Ofir Weber about Complex Barycentri c Coordinates with Applications to Planar Shape Deformation and Interpolat ion at 2010-06-30 11:00:00 DESCRIPTION;LANGUAGE=en-US:Barycentric coordinates are heavily used in com puter graphics applications to generalize a set of given data values. Trad itionally, the coordinates are required to satisfy a number of key propert ies, the first being that they are real. In this work we relax this requir ement, allowing the barycentric coordinates to be complex numbers. This al lows us to generate new families of barycentric coordinates, which have so me powerful advantages over traditional ones and are especially useful for creating detail-preserving planar shape deformations. We first use Cauchy 's theorem to construct three types of complex barycentric coordinates tha t can be used to generate holomorphic functions. Such functions can be int erpreted as conformal planar mappings as long as the derivative of the fun ction doesn't vanish. We then construct another type of complex barycen tric coordinates based on the Hilbert transform. Combined with a novel 2D shape deformation system, we show how to generate "foldovers free" pure co nformal planar deformations. Our system provides the user a large degree o f control over the result. For example, it allows discontinuities at user- specified boundary points, so true "bends" can be introduced into the defo rmation. It also allows the prescription of angular constraints at corners of the target image. Beyond deforming a given shape into a new one at eac h key frame, our method also provides the ability to interpolate between s hapes in a very natural way, such that also the intermediate deformations are conformal. Finally, we show that non-holomorphic complex barycentri c coordinates also exists and may be used to construct planar deformations with exact boundary behavior. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 401 UID:111ec4410400010890 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100630T143000 DTEND;TZID="Asia/Jerusalem":20100630T163000 DTSTAMP;TZID="Asia/Jerusalem":20100630T143000 FREEBUSY;FBTYPE=BUSY:20100630T143000/20100630T163000 SUMMARY;LANGUAGE=en-US:The phd talk by Keren Censor Hillel about Probabili stic Methods in Distributed Computing at 2010-06-30 14:30:00 DESCRIPTION;LANGUAGE=en-US:An inherent characteristic of distributed syste ms is the lack of centralized control, which requires the components to c oordinate their actions. This need is abstracted as the \emph{consensus} problem, in which each process has a binary input and should produce a bi nary output, such that all outputs agree. A difficulty in obtaining conse nsus arises from the possibility of process failures in practical systems . When combined with the lack of timing assumptions in asynchronous syste ms, it renders consensus impossible to solve, as proven by Fischer, Lynch , and Paterson in their fundamental impossibility result, which shows tha t no deterministic algorithm can achieve consensus in an asynchronous sys tem, even if only a single process may fail. Being a cornerstone in distributed computing, much research has been invested in overcoming t his impossibility result. One successful approach is to incorporate rando mization into the computation, allowing the processes to terminate with p robability 1 instead of in every execution, while never violating agreeme nt. This talk will discuss the main challenges in designing random ized consensus algorithms and proving lower bounds. In particular, we pre sent a tight $\Theta(n^2)$ bound for the total step complexity of randomi zed consensus, obtained by improving both known upper and lower bounds. We describe additional problems that arise from the study of randomized c onsensus, including different adversary types, the problem of set agreeme nt, and implementing efficient concurrent data structures. Many open prob lems will be presented throughout the talk. The talk will be self- contained and will assume no prior knowledge in distributed computing. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400010950 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100704T130000 DTEND;TZID="Asia/Jerusalem":20100704T150000 DTSTAMP;TZID="Asia/Jerusalem":20100704T130000 FREEBUSY;FBTYPE=BUSY:20100704T130000/20100704T150000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Amir Vaxman (Computer Scienc e, Technion) about CGGC Seminar: A Multi-Resolution Approach to Heat Kerne ls on Discrete Surfaces at 2010-07-04 13:00:00 DESCRIPTION;LANGUAGE=en-US:Studying the behavior of the heat diffusion pro cess on a manifold is emerging as an important tool for analyzing the ge ometry of the manifold. Unfortunately, the high complexity of the comput ation of the heat kernel - the key to the diffusion process - limits thi s type of analysis to 3D models of modest resolution. We show how to use the unique properties of the heat kernel of a discrete two dimensional manifold to overcome these limitations. Combining a multi-resolution app roach with a novel approximation method for the heat kernel at short tim es results in an efficient and robust algorithm for computing the heat k ernels of detailed models. We show experimentally that our method can ac hieve good approximations in a fraction of the time required by traditio nal algorithms. Finally, we demonstrate how these heat kernels can be us ed to improve a diffusion-based feature extraction algorithm.

Joint work with Mirela Ben Chen (Stanford University) and Craig Gotsman " ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec441040009620 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100705T183000 DTEND;TZID="Asia/Jerusalem":20100705T203000 DTSTAMP;TZID="Asia/Jerusalem":20100705T183000 FREEBUSY;FBTYPE=BUSY:20100705T183000/20100705T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Ram Rachum about Ha ifux, Haifa Linux Club: GarlicSim: An Experimental Tool for Computer Simul ations at 2010-07-05 18:30:00 DESCRIPTION;LANGUAGE=en-US:GarlicSim is an ambitious open-source project i n the field of scientific computing, specifically computer simulations. It attempts to redefine the way that people think about computer simulations , making a new standard for how simulations are created and used. GarlicSi m is a platform for writing, running and analyzing simulations. It is gene ral enough to handle any kind of simulation: Physics, game theory, epidemi c spread, electronics, etc. You may be wondering, what do these simulation s have in common? We'll try to answer this question in this lecture. We wi ll see examples of some simulations in GarlicSim and the tools that it pro vides for running and analyzing them. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400010930 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100711T143000 DTEND;TZID="Asia/Jerusalem":20100711T163000 DTSTAMP;TZID="Asia/Jerusalem":20100711T143000 FREEBUSY;FBTYPE=BUSY:20100711T143000/20100711T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Shay Artzi about Finding and Loc alizing Bugs in Dynamic Web Applications at 2010-07-11 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400010960 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100714T113000 DTEND;TZID="Asia/Jerusalem":20100714T133000 DTSTAMP;TZID="Asia/Jerusalem":20100714T113000 FREEBUSY;FBTYPE=BUSY:20100714T113000/20100714T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Udi Pfeffer (CS, Technion) a bout Pixel Club Seminar: Compressed Sensing for Hyperspectral Imaging at 2 010-07-14 11:30:00 DESCRIPTION;LANGUAGE=en-US:We introduce a system for hyperspectral imaging based on micro-mirror array, that projects subsets of image pixels onto a prism (or diffraction grating), followed by a CCD-type sensor. This s ystem allows generalized sampling schemes including compressed sensing. We acquire only a fraction of the samples that are required to obtain th e full-resolution signal (hyperspectral cube in our case), and by means of non-linear optimization recover the underlying signal. We use a pri or knowledge about the signal sparsity in some dictionary, and its limit ed total variation. We also introduce additional measurements of full-re solution image using small number of filters similar to RGB. As a result , we obtain a feasible system for hyperspectral imaging that enables fas ter acquisition compared to traditional sampling systems. We present sen sitivity analysis of our system to dark and shot noise.

*M .Sc thesis seminar under the supervision of Prof. Ehud Rivlin and Dr. M ichael Zibulevsky. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400010970 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100726T183000 DTEND;TZID="Asia/Jerusalem":20100726T203000 DTSTAMP;TZID="Asia/Jerusalem":20100726T183000 FREEBUSY;FBTYPE=BUSY:20100726T183000/20100726T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Shachar Raindel (EE , Technion) about Haifux, Haifa Linux Club: Valgrind - From Magic to Scien ce at 2010-07-26 18:30:00 DESCRIPTION;LANGUAGE=en-US:Valgrind is an extremely powerful development t ool, which detects and pin-points many common programming errors when runn ing a program. Some examples for the problems it handles are many kinds of buffer overflows, using uninitialized data, accessing freed memory and me mory leaks. Valgrind does that without need for recompilation of the progr am, simply by adding "valgrind" to the beginning of the command line. To an outside observer, this could seem like magic. In this lecture, we will take a peek under the hood of valgrind, and learn how the magic is do ne. I will also briefly introduce other, less known features of valgrind. For more information about valgrind, see Wikipedia and link above. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400010980 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100728T113000 DTEND;TZID="Asia/Jerusalem":20100728T133000 DTSTAMP;TZID="Asia/Jerusalem":20100728T113000 FREEBUSY;FBTYPE=BUSY:20100728T113000/20100728T133000 SUMMARY;LANGUAGE=en-US:The msc talk by Shachar Shem-Tov about Topics in ov er-parametrized based variational methods at 2010-07-28 11:30:00 DESCRIPTION;LANGUAGE=en-US:We discuss a variational methodology, which inv olves locally modeling of data from noisy samples, combined with global m odel parameter regularization. We show that this methodology encompasses many previously proposed algorithms, from the celebrated moving least squ ares methods to the globally optimal over-parametrization methods recentl y published for smoothing and optic flow estimation. However, the unified look at the range of problems and methods previously considered also sug gests a wealth of novel global functionals and local modeling possibiliti es. Specifically, we show that a new non-local variational functional pro vided by this methodology greatly improves robustness and accuracy in loc al model recovery compared to previous methods. The proposed methodology may be viewed as a basis for a general framework for addressing a variet y of common problem domains in signal and image processing and analysis, such as denoising, adaptive smoothing, reconstruction and segmentation. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400013680 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100802T183000 DTEND;TZID="Asia/Jerusalem":20100802T203000 DTSTAMP;TZID="Asia/Jerusalem":20100802T183000 FREEBUSY;FBTYPE=BUSY:20100802T183000/20100802T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Yair Even Zohar (Te chnion) about Haifux, Haifa Linux Club: Zemereshet - An Emergency Project for the Rescue and Electronic Documentation of Early Hebrew Music at 2010- 08-02 18:30:00 DESCRIPTION;LANGUAGE=en-US:Zemereshet is an Emergency Project for the Resc ue and Electronic Documentation of Early Hebrew Music. It is a volunteer a ssociation of enthusiasts who have assumed the mission of preserving the s ongs that were written and sung in Hebrew from the beginning of the Zionis t movement until the establishment of the State of Israel. To that end, th ey established a website with thousands of recordings, lyrics, and complem entary information.

Yair Even-Zohar's talk will present t he project and its various aspects, including copyrights and legal issues, as well as technical issues behind the scenes. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400011010 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100804T113000 DTEND;TZID="Asia/Jerusalem":20100804T133000 DTSTAMP;TZID="Asia/Jerusalem":20100804T113000 FREEBUSY;FBTYPE=BUSY:20100804T113000/20100804T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Sam Hasinoff (CSAIL, MIT) ab out Pixel Club Seminar: Rich Photography on a Budget at 2010-08-04 1 1:30:00 DESCRIPTION;LANGUAGE=en-US:Computation is playing an increasingly central role in how we capture and process our images, opening up richer forms of photography that go beyond conventional imaging. Recent examples include m erging multiple shots to obtain seamless panoramas, 3D shape, deeper focus , or a wider range of tones. In this talk, I will argue that the future of photography lies in richer capture, paying special attention to our limit ed budget of light, time, and sensor throughput. By analyzing tradeoffs an d limits in imaging, we can develop ways to enrich photography while makin g efficient use of our cameras.

First, I will address the bas ic problem of capturing an in-focus image in a fixed time budget. As our a nalysis shows, the number of shots captured is a crucial determinant of qu ality, and taking this into account places the conventional camera in a su rprisingly favorable light. Second, I will describe how existing cameras c an be used more efficiently to capture scenes with a wide range of tones. By adjusting the camera's amplifier as well as its shutter speed, we can a chieve upto 10x noise reduction in the darkest parts of the scene. Both of these projects demonstrate how computation can enrich photography, while providing significant gains over the state-of-the-art. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400011020 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100810T113000 DTEND;TZID="Asia/Jerusalem":20100810T133000 DTSTAMP;TZID="Asia/Jerusalem":20100810T113000 FREEBUSY;FBTYPE=BUSY:20100810T113000/20100810T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Gil Rivnai (EE, Technion) a bout Pixel Club Seminar: Measurement and Analysis the Biological Retinal R esponse, and Adaptation to Optical Stimulus at 2010-08-10 11:30:00 DESCRIPTION;LANGUAGE=en-US:The retina is the first nerve system tissue tha t processes the visual information at the biological visual system. At thi s research, lab experiment performed to measure the response of the retina l response, the examination was perform from spatial point of view, by sim ultaneous recording from many electrodes that recording the ganglion cell response from different points on the retina. First found the optimal puls e duration that led to the maximal response, and the adaptation time const ant of the retina. In additional the response delay of the retina for th e optimal stimulus. Series of different pulse duration and the recovery t ime duration stimulate the retina, to conclude the effect of those paramet ers on the retinal response and the adaptation characteristic. the retinal response to un uniform chess board stimulus signal examined, while focuse d the response at different area of the retina (off process, on process, e dges). Also at this series of experiments the same parameters of the respo nse examined, to determined the effect of the spatial and time stimulus ch aracteristic on those parameters. We introduce the data processing and dia gnostic methods of the huge amount of data that recorded, led to visual pr esentation of the results and the conclusions while overcoming low SNR and emphasize the retinal response. The research conclusion is that biologica l retinal response characterization could use for computerized bionic visi on by diagnose and process the retinal response and convert it to a image, and assist the research that perform to develop vision systems.
< br> M.Sc. thesis under the supervision of Dr. Moshe Porat for EE and Pro f. Ido Prelman from Medicine. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400011000 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100816T183000 DTEND;TZID="Asia/Jerusalem":20100816T203000 DTSTAMP;TZID="Asia/Jerusalem":20100816T183000 FREEBUSY;FBTYPE=BUSY:20100816T183000/20100816T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Tzafrir Rehan about Haifux, Haifa Linux Club: HTML5 - The Next Generations of the Web at 2010 -08-16 18:30:00 DESCRIPTION;LANGUAGE=en-US:Since 2004, the Web Hypertext Application Techn ology Working Group (WHATWG) has been working on the next generation of HT ML, the markup that makes up web documents.

In this talk, we will dive into some of the new elements defined by HTML5, see examples of how behavior that used to require third party software can now be impleme nted using web standards, discuss the evolution of HTML, and explore the o ptions of developing fully featured applications on top of the web applica tion stack that is currently being formed. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400011030 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100829T143000 DTEND;TZID="Asia/Jerusalem":20100829T163000 DTSTAMP;TZID="Asia/Jerusalem":20100829T143000 FREEBUSY;FBTYPE=BUSY:20100829T143000/20100829T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Eldad Haber SPECIAL LECTURE abou t An effective method for parameter estimation with PDE constraints with m ultiple right hand sides at 2010-08-29 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400010990 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100830T100000 DTEND;TZID="Asia/Jerusalem":20100830T120000 DTSTAMP;TZID="Asia/Jerusalem":20100830T100000 FREEBUSY;FBTYPE=BUSY:20100830T100000/20100830T120000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by John Keyser (Computer Scienc e and Engineering at Texas A&M University) about Pixel Club Seminar: Recon struction of the Brain Using High-Throughput Microscopy: Imaging, Segmenta tion, and Visualization at 2010-08-30 10:00:00 DESCRIPTION;LANGUAGE=en-US:Recent advances in microscopy have enabled the collection of large amounts of microscopic data at rapid rates. The size a nd rate of acquisition of these "high throughput" microscopic data sets pr esent new opportunities in understanding the structure and function of org ans, but also present new challenges in the processing, analysis, and visu alization of the data. This talk will describe some of the challenges and opportunities coming from an effort to reconstruct the mouse brain. This t alk will focus particularly on work being done in the Brain Networks Labor atory (BNL) at Texas A&M University. The BNL works with data scanned from a Knife-Edge Scanning Microscope (KESM) developed in our lab. Our particul ar emphasis has been the scanning and reconstruction of neuronal and micro vascular data from mouse brain. This talk will present an overview of the KESM and other high-throughput techniques, along with the challenges in im age processing, segmentation, reconstruction, and visualization. Of partic ular interest are long, thin structures encountered in the brain, such as neuronal processes and microvasculature. We will discuss some of the algor ithms used to segment and reconstruct this data, along with methods used t o provide useful visualizations to collaborators interested in gaining bet ter understanding of the data. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400011040 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100902T143000 DTEND;TZID="Asia/Jerusalem":20100902T163000 DTSTAMP;TZID="Asia/Jerusalem":20100902T143000 FREEBUSY;FBTYPE=BUSY:20100902T143000/20100902T163000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Martin Ehler (Departments of Mathematics and Earth and Ocean Science, University of British Columbi a) about Pixel Club Seminar: A Parallel Algorithm for Sparse Demixing - Applied to Retinal Imaging and Remote Sensing at 2010-09-02 14:30:00 DESCRIPTION;LANGUAGE=en-US:The talk addresses the analysis of multi-spectr al retinal image sets and hyperspectral satellite images. Pixels in such image sets represent mixtures of pure substances, called endmembers. Th e linear mixture model for hyperspectral imaging assumes that each pixel is a linear combination of the spectra of a collection of endmembers. T o demix each pixel, these endmembers must be identified. A novel paralle l algorithm to extract endmembers and demix each pixel is presented.

Typical algorithms for endmember extraction do not use that each pi xel is a combination of only few endmembers. Unlike others, our algorith m is based on the sparse representation of each pixel. The proposed algo rithm solves a series of independent minimization problems with sparsity constraints and then identifies endmembers according to a ranking of th e computed coefficients. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011050 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100914T113000 DTEND;TZID="Asia/Jerusalem":20100914T133000 DTSTAMP;TZID="Asia/Jerusalem":20100914T113000 FREEBUSY;FBTYPE=BUSY:20100914T113000/20100914T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Matan Protter (CS, Technion) about Pixel Club Seminar: Image Sequence Processing Without Motion Estima tion at 2010-09-14 11:30:00 DESCRIPTION;LANGUAGE=en-US:YouTube movies, live streaming, TV broadcast, c onference calls and more - there is no doubt that video sequences are abun dant and in everyday use. However, the quality of these videos is rarely s atisfactory. This may be the result of network limitations, low-quality im aging devices and more.

Improving the quality of videos has been long studied by the research community. Rather than independently im proving the quality of each image in the sequence, it has been observed th at temporal consideration can greatly improve the outcome, due to details being shown in several consecutive images. In order to take advantage of t his observation, existing research till recently focused on estimating the motion between the images. However, motion estimation is quite a complex task, and might result in errors, which degrade the overall quality of the video enhancement process.

In the last couple of years, th ere has been a trend of applying temporal filtering to videos while bypass ing the need for motion estimation. In this talk, we will present two diff erent tasks able to rely on this new approach. We will start by briefly di scussing the removal of noise from video, using sparse and redundant repre sentations modeling of images.

Using the intuition gained f rom this problem, we will turn to one of the holy grails of video processi ng - super resolution. Super-resolution (SR) deals with the ability to mer ge a sequence of low-quality images into one (or a sequence of) high-quali ty image, having better optical resolution and with new details visible. U nfortunately, though existing for over 20 years, SR has suffered greatly f rom the need for extremely accurate motion estimation. SR has been applica ble to only a small portion of the sequences requiring it, as in most sequ ences, the motion patterns are too complex to be accurately estimated by e xisting means.

In the lion's share of this talk we will sho w that it is possible to circumvent the need for accurate deterministic mo tion estimation by replacing it with crude, probabilistic motion estimatio n. This allows us to improve the quality of sequences previously considere d out of the realm of super-resolution. We show some examples, both synthe tic and on actual broadcast data, demonstrating the ability of this approa ch.

This lecture summarizes portions of a PhD. research by Matan Protter, under the supervision of Prof. Michael Elad. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011060 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100915T133000 DTEND;TZID="Asia/Jerusalem":20100915T153000 DTSTAMP;TZID="Asia/Jerusalem":20100915T133000 FREEBUSY;FBTYPE=BUSY:20100915T133000/20100915T153000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Miriam Ragle Aure (Department of Genetics, Institute for Cancer Research, Oslo University an d Hospital Radiumhospitalet, Oslo, Norway ) and Ole-Christian Lingjær de (Centre for Cancer Biomedicine, Un about Bioinformatics Forum: Towards The Systems Biology of Breast Cancer: Exploiting Multilevel Molecular Data of High Dimension at 2010-09-15 13:30:00 DESCRIPTION;LANGUAGE=en-US:Cancer is a worldwide burden with several milli on deaths annually and the situation is set to worsen globally as the po pulation ages, with a projected increase of 45% to 2030 according to the WHO. Current cancer management is mainly focused on intervention after tumors have been detected; however, there is a drive towards very early detection and intervention to reduce the risk of metastatic disease. Nov el methodology is emerging to allow modulation of tumor progression thro ugh targeting of molecular pathways in particular molecular subclasses. A major challenge for cancer prevention and treatment success is that th e development and progression are complex biological/evolutionary proces ses, involving tumors that are composite organ systems with dynamical ge nomes shaped by gene aberrations, epigenetic changes, cellular biological context, characteristics specific to the individual patient, and environ mental influences. Conventional biological and statistical approaches ar e not designed to unravel the complicated interactions involved in cellu lar functioning. In particular, biological interactions will increasingl y have to be sought at and between several systemic levels. Failure to c apture such connections may seriously compromise our ability to distingu ish between normal variation and disease-related perturbations in a path way. In a systems biology perspective, tumor development and progression depends on an interplay between different endogenous processes at vario us levels, as well as external stimuli.

Ultimately, we would like to be able to predict the effect of any given perturbation of this system. While this is a very ambitious goal,the perspective of systems biology is nevertheless of great importance,as we need to move towards m ore systematic studies of complex interactions in order to better unders tand the entirety of the processes. A substantial gap exists between the abundance and complexity of the data typically available for a patient/tu mor and the sophistication of the statistical and computational tools available to handle and interpret this information. High-throughput molecu lar data at various levels, including DNA, mRNA, miRNA, protein and meth ylation are being generated on a steadily increasing number of breast ca ncer patients. Several molecular markers have been identified and are us ed to develop profiles associated with tumor aggressiveness, response to therapy and patient outcome. Pooling datasets, combining profiles at va rious levels, and analyzing the data in a compendium rather than in isol ation can lead to more reliable molecular signatures and thereby more sp ecific diagnosis and treatment of breast cancer patients. It is usually not feasible to obtain data at all systemic levels on all patients and on normal as well as tumor tissue. More typically, some measurements are available on one group of individuals and other measurements are availa ble on another group. Sometimes, groups partially overlap so that more m easurements are available on a smaller subset of individuals. A major ch allenge in the analysis of multilevel molecular data is to develop methods and algorithms that allow joint analysis of such partially overlapping data sets. In this talk, we will discuss some of the challenges in this context and provide own examples of how these may be approached. We will also share some speculations concerning how to handle such problems in the future. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400011070 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100921T113000 DTEND;TZID="Asia/Jerusalem":20100921T133000 DTSTAMP;TZID="Asia/Jerusalem":20100921T113000 FREEBUSY;FBTYPE=BUSY:20100921T113000/20100921T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Ayelet Dominitz (EE, Technio n) about Pixel Club Seminar: Surface and Volume Mapping via Mass Transport at 2010-09-21 11:30:00 DESCRIPTION;LANGUAGE=en-US:In this talk, we present novel approaches for m apping surfaces and volumes to corresponding canonical domains with the sa me topology. In general we wish to find a bijection of the sampled surface or volume lying in 3D space with a simpler model, that respectively prese rves the area or volume distribution and minimally distort the local geome try. Many of the operations we wish to perform on the discrete surface o r volume, for processing, analysis or visualization, may be greatly simpli fied if we perform them in a corresponding canonical domain with the same topology. For example, the surface of the brain is a topological sphere, b ut is highly convoluted. If one can find a good bijection of the brain sur face onto the sphere, this would simplify many types of visualization and analysis tasks.

Our surface mapping (parameterization) method aims at being optimal for many applications by creating a map to the s phere that minimizes the geometrical distortion while being area preser ving. Our approach is based on the technique of optimal mass transport (OMT), also known as the "earth mover's problem.'' In our work we derived a novel formulation of the gradient descent minimizing flow for OMT proble m. As well as being very useful for the efficient computation of surface p arameterization, we believe that this flow is also quite interesting from a theoretical point of view, because it enables us to solve the optimal ma ss transport problem over any Riemannian manifold.

Our volume mapping method aims at being volume preserving. Using this method we can map any given solid to the cube. This mapping may be practical for simplif ying calculations such as constrain analysis. In addition, the mapping pro cess we use not only maps any given solid to a cube, but also generates a regular hexahedral mesh of the given solid, with very small variance of th e volume of the hexahedral elements. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400011080 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20100921T143000 DTEND;TZID="Asia/Jerusalem":20100921T163000 DTSTAMP;TZID="Asia/Jerusalem":20100921T143000 FREEBUSY;FBTYPE=BUSY:20100921T143000/20100921T163000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Michael Shamis (EE, Technion ) about Pixel Club Seminar: Blind Source Separation of Instantaneous at 20 10-09-21 14:30:00 DESCRIPTION;LANGUAGE=en-US:Blind source separation of images and voice sig nals is a well known and well studied subject. Solutions for this problem have various applications, such as separation of voices of multiple speake rs in the same room, denoising, separation of reflections superimposed on images, and more. Classical time/position invariant Blind Source Separa tion is usually solved using Independent Component Analysis (ICA), which t ries to find statistically independent signals as a linear combination of the mixed signals, or by using Sparse Component Analysis (SCA) which estim ates the mixing matrix by analyzing the geometry of the problem and uses s catter plots of the mixed signals to estimate co-linear centroids of the s cattered data, where each centroid corresponds to a column in the mixing m atrix. Most of the studies in the field assume time/position invariant signal combinations, although many real problems are not such. Recently, i n his PhD thesis, Ran Kaftory has proposed an extension of the SCA method to solve multiple families of the time/position varying problems. He has s hown that for instantaneous time/position varying mixtures, the problem of lines estimation transforms to estimation of non linear curves. In this r esearch we explore the separation of instantaneous time/position varying m ixtures for which the parametric structure of the mixtures family is known a priori .We show that the geometric approach can also be viewed as Maxim um Likelihood (ML) problem when spasification is applied to the mixed sign als. We propose a multi-staged SCA algorithm for separation of time/positi on invariant mixtures and extend the solution to a subset of time/position varying mixtures where the reconstruction is performed by curve fitting t echniques and nearest neighbor clustering. In addition to the geometric approach, we extend the well known technique of ICA, through ML , to the case of time/position varying instantaneous mixtures. We show that ML appr oach to the time/position varying separation problem can be developed from an information theory perspective as a joint entropy minimization of the unmixed signals. We prove that although the problem is non-convex, and may need non-linear optimization techniques, under certain conditions the cor rect signals separation constitutes a global maximum of the ML optimizatio n problem.

We conclude by showing that the ML approach provi des promising results, but due to the non-linear nature of the problem, it s optimization is challenging and SCA-based approach can be used as a comp limentary technique to circumvent some of the difficulties originating fro m the non linearities of the problem. *PhD thesis under the supervis ion of Prof. Yehoshua Y. Zeevi. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400011100 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101004T183000 DTEND;TZID="Asia/Jerusalem":20101004T203000 DTSTAMP;TZID="Asia/Jerusalem":20101004T183000 FREEBUSY;FBTYPE=BUSY:20101004T183000/20101004T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Orr Dunkelman (Weiz mann Institute of Science) about Haifux, Haifa Linux Club: The Hitchhik ers' Guide to the SHA-3 Competition at 2010-10-04 18:30:00 DESCRIPTION;LANGUAGE=en-US:Following the advances in the cryptanalysis of hash functions in the last years, and especially the alarming results o n SHA1 and Merkle-Damgard hash functions, NIST has started a cryptograp hic competition to offer secure and fast hash function standard, to be named SHA3. The competition, currently at its second phase (with 14 rem aining candidates), is the focus of many efforts in cryptanalysis and i mplementation.

In this talk, we shall cover some of the aspec ts of this competition, starting from the reasons that led to it, throu gh the first phase, the 14 candidates that were chosen for the second r ound, their status (both security-wise and performance-wise), and of co urse, what is left to be done, so the outcome of the process will be th e best possible SHA3. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400011120 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101005T113000 DTEND;TZID="Asia/Jerusalem":20101005T133000 DTSTAMP;TZID="Asia/Jerusalem":20101005T113000 FREEBUSY;FBTYPE=BUSY:20101005T113000/20101005T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Ron Rubisnstein (CS, Technio n) about Pixel Club Seminar: Sparsity-Based Signal Models and the Sparse K -SVD Algorithm at 2010-10-05 11:30:00 DESCRIPTION;LANGUAGE=en-US:Signal models are used for a wide array of sign al and image processing tasks – from deconvolution, denoising, and inter polation to source separation, super-resolution, and compression. One of t he most common modeling approaches utilizes a dictionary of atomic signals , describing the set of elementary behaviors observed in the signals of in terest. The dictionary is used either as an analysis operator, measuring t he inner-products of the atoms and the input signal, or as a synthesis ope rator, reproducing the input signal as a linear combination of the atoms. In both cases, the driving force of the model is sparsity, which requires that the representation coefficients be mostly close or equal to zero. In this talk we begin by presenting the analysis versus synthesis question, w hich has been gaining interest in recent years. We provide some surprising theoretical results which show that, in contrast to common belief, the tw o approaches may substantially differ, and discuss the implications of the se results. In practice, the success of these models depends on the choice of the sparsifying dictionary. Most dictionaries emerge from one of two s ources – either a mathematical approximation of the signal data, which l eads to structured and efficient dictionaries, or example-based training, which produces adaptive and highly tuned dictionaries. In the main part of this talk, we present the sparse dictionary structure, which aims to narr ow the gap between these two options. Focusing on the synthesis case, we d escribe the new structure and its advantages, and discuss the Sparse K-SVD algorithm which learns it from examples. We discuss the application of th e new structure to volume denoising and image compression, where we demons trate, for the first time, feasibility of a fully data-adaptive scheme for general-purpose compression. Several additional applications are briefly described.

This lecture summarizes portions of a PhD. research by Ron Rubisnstein, under the supervision of Prof. Michael Elad. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011110 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101019T113030 DTEND;TZID="Asia/Jerusalem":20101019T133030 DTSTAMP;TZID="Asia/Jerusalem":20101019T113030 FREEBUSY;FBTYPE=BUSY:20101019T113030/20101019T133030 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Soraia Raupp Musse (PUCRS - Computing Science Department) about Pixel Club Seminar: Integrating Com puter Graphics and Computer Vision in Virtual Humans Projects at 2010-1 0-19 11:30:30 DESCRIPTION;LANGUAGE=en-US:In this talk I'll present some research topics supervised by me during the last 10 years in VHLab, mainly in the conte xt of Virtual Humans, and, more recently, in computer vision (CV) integ rated with computer graphics (CG). Firstly, I'll discuss topics focused on virtual humans simulation and open problems. Then, integration betw een CG and CV is presented, applied in practical and relevant areas. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011170 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101019T143000 DTEND;TZID="Asia/Jerusalem":20101019T163000 DTSTAMP;TZID="Asia/Jerusalem":20101019T143000 FREEBUSY;FBTYPE=BUSY:20101019T143000/20101019T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Remi Gribonval about Should pena lized least squares regression be interpreted as Maximum A Posteriori esti mation? Subtitle: Reading into the map: how mean is it ? at 2010-10-19 14: 30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011130 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101019T143030 DTEND;TZID="Asia/Jerusalem":20101019T163030 DTSTAMP;TZID="Asia/Jerusalem":20101019T143030 FREEBUSY;FBTYPE=BUSY:20101019T143030/20101019T163030 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Remi Gribonval (INRIA) about Pixel Club Seminar: Reading into the MAP: how mean is it ? at 201 0-10-19 14:30:30 DESCRIPTION;LANGUAGE=en-US: Should penalized least squares regression be interpreted as Maximum A Posteriori estimation? Penalized least sq uares regression is often used for signal denoising and inverse problems , and is commonly interpreted in a Bayesian framework as a Maximum A Pos teriori (MAP) estimator, the penalty function being the negative logarit hm of the prior. For example, the widely used quadratic program (with an $\ell^1$ penalty) associated to the LASSO / Basis Pursuit Denoising is very often considered as the MAP under a Laplacian prior in the context of additive white Gaussian noise (AWGN) reduction. The objective of this talk is to highlight the fact that, while this is {\em one} possible Ba yesian interpretation, there can be other equally acceptable Bayesian in terpretations. Therefore, solving a penalized least squares regression p roblem with penalty $\phi(x)$ need not be interpreted as assuming a prior $C\cdot \exp(-\phi(x))$ and using the MAP estimator. In particular, I wi ll show that for {\em any} prior $p_X(x)$, the conditional mean estimato r can be interpreted as a MAP with some prior $C \cdot \exp(-\phi(x))$. Vice-versa, for {\em certain} penalties $\phi(x)$, the solution of the penalized least squares problem is indeed the {\em conditional mean}, wit h a certain prior $p_X(x)$. In general we have $p_X(x) \neq C \cdot \e xp(-\phi(x))$.

If time allows I will also discuss recent j oint work with Volkan Cevher (EPFL) and Mike Davies (University of Edinb urgh) on "compressible priors", in connection with sparse regularization in linear inverse problems such as compressed sensing. I will show in p articular that Laplace distributed vectors cannot be considered as typic ally "compressible": they are not sufficiently well approximated by spa rse vectors to be recovered from a low-dimensional random projection by, e.g., L1 minimization. This may be somewhat of a surprise considering that L1 minimization is associated to MAP estimation under the Laplace p rior! ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011180 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101020T123000 DTEND;TZID="Asia/Jerusalem":20101020T143000 DTSTAMP;TZID="Asia/Jerusalem":20101020T123000 FREEBUSY;FBTYPE=BUSY:20101020T123000/20101020T143000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Yossi Gil (CS, Technion) a bout Pixel Club Seminar: Schizophrenic(*) Perspective on Software Testing at 2010-10-20 12:30:00 DESCRIPTION;LANGUAGE=en-US:In this talk, I will present my personal perspe ctive on software testing and quality, a perspective which is torn between being an academic interested in programming languages, and my earthly exp erience, working for the last two years as a programmer in the trenches wi th Google and IBM.

This talk is a re-run of an invited talk at HVC'10 conference.

(*) As will be explained in my talk, the term "schizophrenic" is used here in a non-polit ically correct-, non-professional and inaccurate- sense, a sense which cor responds to daily meaning of the word. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 2 UID:111ec4410400011150 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101020T140000 DTEND;TZID="Asia/Jerusalem":20101020T160000 DTSTAMP;TZID="Asia/Jerusalem":20101020T140000 FREEBUSY;FBTYPE=BUSY:20101020T140000/20101020T160000 SUMMARY;LANGUAGE=en-US:The phd talk by Ilia Averbouch about Completeness a nd Universality Properties of Graph Invariants and Graph Polynomials at 20 10-10-20 14:00:00 DESCRIPTION;LANGUAGE=en-US:Graph polynomials are powerful and well-develop ed tools to express graph parameters. Usually graph polynomials are compa red to each other by ad-hoc means allowing to decide whether a newly defi ned graph polynomial generalizes (or is generalized) by another one. We study their distinctive power and introduce the notions of $dp$-comple teness and universality of graph polynomials in order to formalize depend encies between them. Many known graph polynomials satisfy linear recurr ence relations with respect to some set of edge- or vertex-elimination op erations. Inspired by the work of Brylawski and Oxley on the Tutte polyno mial, we define several classes of graph polynomials according to their recurrence relations, and prove $dp$-completeness and universality result s for those classes. The talk is self-contained and does not require an y prior knowledge on graph polynomials. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400011140 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101025T183000 DTEND;TZID="Asia/Jerusalem":20101025T203000 DTSTAMP;TZID="Asia/Jerusalem":20101025T183000 FREEBUSY;FBTYPE=BUSY:20101025T183000/20101025T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Orr Dunkelman (Weiz mann Institute of Science) about Haifux, Haifa Linux Club: Secure File Sys tems at 2010-10-25 18:30:00 DESCRIPTION;LANGUAGE=en-US:In this talk I shall cover two concepts related to protecting your information, cryptographic file system and steganograp hic file system. A cryptographic file system is a file system where the da ta stored is encrypted by the operating system (thus, protecting the confi dentiality of the information in case the hard disk is taken away). A steg anographic file system is a file system which is hidden from prying eyes, namely, that a person holding the hard disk cannot even find the files.

The talk will describe some of the technical aspects of these file systems, including a bit of the concepts of how they work, and the pr oblems they may cause. Then, as much as possible, we shall discuss their i mplementations in Linux.

The talk will not assume crpytograp hic knowledge, even though, it won't hurt you to know a bit before hand. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400011090 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101027T133030 DTEND;TZID="Asia/Jerusalem":20101027T153030 DTSTAMP;TZID="Asia/Jerusalem":20101027T133030 FREEBUSY;FBTYPE=BUSY:20101027T133030/20101027T153030 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Michal Genussov (EE, Technio n) about Pixel Club Seminar: Transcription and Classification of Audio dat a by Sparse Representations and Geometric Methods at 2010-10-27 13:3 0:30 DESCRIPTION;LANGUAGE=en-US:Transcription and Classification of Audio Data by sparse Representations and Geometric Methods ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400011160 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101027T140000 DTEND;TZID="Asia/Jerusalem":20101027T160000 DTSTAMP;TZID="Asia/Jerusalem":20101027T140000 FREEBUSY;FBTYPE=BUSY:20101027T140000/20101027T160000 SUMMARY;LANGUAGE=en-US:The msc talk by Ilia Nudelman about Dependent UFP O n a Shared Channel With Application to a Network Centric Operation at 2010 -10-27 14:00:00 DESCRIPTION;LANGUAGE=en-US:In this talk we generalize the well-known Unspl ittable Flow Problem (UFP) and show its applicability to the emerging N etwork Centric Operations (NCO) concept. While UFP's goal is to maximiz e the profit gained by accommodating independent flows, our generalizat ion considers dependent flows. In the generalized problem, referred to as the Dependent Unsplittable Flow Problem (D-UFP), the profit from del ivering two flows is not necessarily equal to the sum of their profits. We show that in a single channel NCO, D-UFP can be approximated as the Qu adratic Knapsack Problem (QKP). We present new algorithms for QKP, and examine their performance in typical NCO scenarios. Our simulation stud y shows that when the dependency between flows is taken into account, t he performance of a typical NCO scenario can increase by 10-100%. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400011190 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101102T113000 DTEND;TZID="Asia/Jerusalem":20101102T133000 DTSTAMP;TZID="Asia/Jerusalem":20101102T113000 FREEBUSY;FBTYPE=BUSY:20101102T113000/20101102T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Vadim Indelman (Aerospace En gineering, Technion) about Pixel Club Seminar: Navigation Performance Enha ncement using Online Mosaicking at 2010-11-02 11:30:00 DESCRIPTION;LANGUAGE=en-US:Since the Global Positioning System (GPS) was e stablished in the 1970s, navigation has become a much easier task. Indeed, the majority of navigation systems rely on the GPS signal for correcting the developing dead reckoning errors. However, GPS is unavailable or unrel iable indoors, underwater, in urban environments, and on other planets. In these scenarios, one must use alternative techniques for updating the Ine rtial Navigation System (INS), or any other dead reckoning mechanism being used. This research presents algorithms that use imagery captured durin g motion and mosaics constructed based on these images, for performing nav igation aiding. The vehicle is assumed to be equipped with an INS and a si ngle camera. No other sensors or a-priori information are required.

First, an algorithm for navigation aiding will be presented which c ouples between the process of online mosaic construction and the process o f a gimbaled camera scanning. Improved navigation precision is obtained, c ompared to a basic method for image-aided navigation, when operating in ch allenging scenarios that include a narrow field-of-view camera observing l ow-texture scenes.

Next, a new set of constraints will be der ived by analyzing a general static scene, observed from three different vi ews. A formulation of an implicit extended Kalman filter will be presented for fusing these constraints with an INS. The developed method reduces th e navigation errors, including position and velocity errors in all axes wi thout explicit scale recovery. In particular, the method efficiently handl es scenarios in which the same scene is observed several times.

A natural extension of this method is to consider a group of cooperativ e platforms that exchange imagery and navigation information. The three-vi ew constraints are formulated whenever the same scene is observed by diffe rent platforms, not necessarily at the same time. Since the navigation inf ormation involved in the three-view constraints may be statistically depen dant, the appropriate correlation terms should be known in order to obtain consistent information fusion. A new graph-theory-based technique will be presented for on-demand calculation of the required correlation terms for a general multi-platform measurement model.

Both simulation and experimental results will be presented validating the algorithm perfor mance.

Ph.D. seminar under the supervision of Prof. Pini Gurf il, faculty of Aerospace Engineering, Prof. Ehud Rivlin, department of Com puter Science, and Dr. Hector Rotstein, Rafael. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011230 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101102T143000 DTEND;TZID="Asia/Jerusalem":20101102T163000 DTSTAMP;TZID="Asia/Jerusalem":20101102T143000 FREEBUSY;FBTYPE=BUSY:20101102T143000/20101102T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Marius Ungarish about Gravity cu rrents as a test case: do simple mathematical models produce good physical insights, or do good insights generate powerful models? at 2010-11-02 14: 30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011210 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101108T183000 DTEND;TZID="Asia/Jerusalem":20101108T203000 DTSTAMP;TZID="Asia/Jerusalem":20101108T183000 FREEBUSY;FBTYPE=BUSY:20101108T183000/20101108T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Shahar Dag (SSDL, C S, Technion) about Haifux, Haifa Linux Club: SSDL & Linux – a Love story or a Battle field? How to Change to Linux and Survive (Give or Take) at 2 010-11-08 18:30:00 DESCRIPTION;LANGUAGE=en-US:SSDL & Linux – a Love story or a Battle field ?

In the talk I will describe the process of moving (almost) all the computers of SSDL (the System & Software Development Laboratory at the Computer science faculty - Technion) from Windows to Linux.

I will talk about the rezones for the move and about the benefits & dr awback of Linux from the administrator (that is me) point of view and from the student's point of view.

It will be great if we will hav e some kind of open discussion about the selection of the best OS for (hom e, academic, scientific …) usage. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400011240 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101109T143000 DTEND;TZID="Asia/Jerusalem":20101109T163000 DTSTAMP;TZID="Asia/Jerusalem":20101109T143000 FREEBUSY;FBTYPE=BUSY:20101109T143000/20101109T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Eli Ben-Sasson about Size-space tradeoffs in propositional proof complexity at 2010-11-09 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011220 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101110T123000 DTEND;TZID="Asia/Jerusalem":20101110T143000 DTSTAMP;TZID="Asia/Jerusalem":20101110T123000 FREEBUSY;FBTYPE=BUSY:20101110T123000/20101110T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Zohar Karnin (CS, Techni on) about Theory Seminar: Deterministic Construction of a high dimensional ell-p section in ell-1^n for any p<2 (and its implications in compressed sensing) at 2010-11-10 12:30:00 DESCRIPTION;LANGUAGE=en-US:For any $00$, we give an efficient deterministic construction of a linear subspace $V \subseteq \R^n$, of dimension $(1-\epsilon)n$ in which the $\ell_p$ and $\ell_r$ norms are the same up to a multiplicative factor of $\poly(\epsilon^{-1 })$ (after the correct normalization). As a corollary we get a determini stic compressed sensing algorithm (Base Pursuit) for a new range of para meters. In particular, for any constant $\epsilon>0$ and $p<2$, we obtai n a linear operator $A:\R^n \rightarrow \R^{\epsilon n}$ with the $\ell_ 1/\ell_p$ guarantee for $(n \cdot \poly(\epsilon))$-sparse vectors. Name ly, let $x$ be a vector in $\R^n$ whose $\ell_1$ distance from a $k$-s parse vector (for some $k=n \cdot \poly(\epsilon)$) is $\delta$. The alg orithm, given $Ax$ as input, outputs an $n$ dimensional vector $y$ such that $||x-y||_p \leq \delta k^{1/p-1}$. In particular this gives a weak form of the $\ell_2/\ell_1$ guarantee.

Our construction has the additional benefit that when viewed as a matrix, $A$ has at most $O (1)$ non-zero entries in each row. As a result, both the encoding (compu ting $Ax$) and decoding (retrieving $x$ from $Ax$) can be computed effic iently. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011290 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101116T113000 DTEND;TZID="Asia/Jerusalem":20101116T133000 DTSTAMP;TZID="Asia/Jerusalem":20101116T113000 FREEBUSY;FBTYPE=BUSY:20101116T113000/20101116T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Yosef Yomdin (Math and CS, W eizmann Institute of Science) about Pixel Club Seminar: Representation, P rocessing and Animation of Images via Geometric Models at 2010-11-16 1 1:30:00 DESCRIPTION;LANGUAGE=en-US:In this talk I plan to discuss some results on image representation and processing via geometric models, in two extreme s cales:

1. The finest scale representation is usually called "ve ctorization". Here the models are analytic aggregates mainly constructed f rom edges and ridges, equipped with their "extended color profiles". We pr ovide some experimental data stressing the importance of accurate and flex ible ``color profiles" of edges and ridges in our visual perception. Accor dingly, the extended color profiles play the most central role in our vect or representation and in its editing in "vector form". We describe our vec tor format which allows (for the first time, as we believe) for a "visuall y indistinguishable" reconstruction of high-resolution photo-images, and f or their geometric processing.

2. Another extreme is the prob lem of capturing of global characters and objects on the image with global parametric models, like the model of an entire human figure. This is a ve ry important step in "photo-animation". Recently motion data bases have be en used in combination with model fitting in a work of Aachen Computer Vis ion group to provide high-quality animations of certain photo-characters. However, essential difficulties arise when the output 2D animation format is relatively rigid layers based one, like Flash. Another problem appears when the anatomic proportions and the pose of the character on the photo d iffer strongly from that used in the motion data-base. We show in this tal k how to transform properly the data-base 3D animations into a 2D layer-ba sed format, in order to preserve an illusion of a 3D motion, and how to co rrect it in order to make it appropriate for a specific pose and anatomic proportions of the photo-character.

Department of Mathematic s, The Weizmann Institute of Science, Rehovot, Israel. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011280 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101117T123000 DTEND;TZID="Asia/Jerusalem":20101117T143000 DTSTAMP;TZID="Asia/Jerusalem":20101117T123000 FREEBUSY;FBTYPE=BUSY:20101117T123000/20101117T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Oren Ben-Zwi (Technion) about Theory Seminar: Hats Auctions and Derandomization at 2010-11-17 12:3 0:00 DESCRIPTION;LANGUAGE=en-US:Goldberg, Hartline and Wright [SODA 2001] intro duced the notion of 'competitive analysis' into 'auction theory' when deal ing with 'unlimited supply', 'unit demand', 'single item' auctions. They proved that there exists random auctions that guaranty constant 'competiti ve ratio' and that no deterministic auction can guaranty that.

Aggarwal, Fiat, Goldberg, Hartline, Immorlica and Sudan [STOC 2005] show ed that the 2001 conclusion was wrong and that there exist deterministic a symmetric auction with constant competitive ratio. They also show an elega nt derandomization that uses some intermediate 'hat guessing games' and ha ve a factor 4 loss from the random expectation.

We show a new derandomization using the 'Lovasz's local lemma' that eliminate this loss . We also show a specific setting for which there exists an elegant polyno mial deterministic auction accompanied with a matching lower bound. For this we solve a different 'hat guessing game'. This answers an open questi on of Feige [2004].

The talk is based on works done jointly w ith Guy Wolfovitz and Ilan Newman. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011310 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101117T133000 DTEND;TZID="Asia/Jerusalem":20101117T153000 DTSTAMP;TZID="Asia/Jerusalem":20101117T133000 FREEBUSY;FBTYPE=BUSY:20101117T133000/20101117T153000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Noam Shental (CS, The Open University of Israel) about Bioinformatics Forum: Identification of Rare Alleles and their Carriers using Compressed Se(que)nsing at 2010-1 1-17 13:30:00 DESCRIPTION;LANGUAGE=en-US:Identification of rare variants by resequencing is important both for detecting novel variations and for screening indi viduals for known disease alleles. New technologies enable low-cost rese quencing of target regions, although it is still prohibitive to test mor e than a few individuals. We propose a novel pooling design that enables the recovery of novel or known rare alleles and their carriers in group s of individuals. The method is based on a Compressed Sensing (CS) appro ach, which is general, simple and efficient. CS allows the use of generi c algorithmic tools for simultaneous identification of multiple variants and their carriers. We model the experimental procedure and show via co mputer simulations that it enables the recovery of rare alleles and their carriers in larger groups than were possible before. Our approach can al so be combined with barcoding techniques to provide a feasible solution based on current resequencing costs. For example, when targeting a small enough genomic region (~100 bp) and using only ~10 sequencing lanes and ~10 distinct barcodes per lane, one recovers the identity of 4 rare all ele carriers out of a population of over 4000 individuals. We demonstrat e the performance of our approach over several publicly available experi mental data sets, including the 1000 Genomes Pilot 3 study. We believe our approach may significantly improve cost effectiveness in future Gen ome Wide Association Studies, and in screening large DNA cohorts for spe cific risk alleles.

Joint work with Amnon Amir from the Weizm ann Institute of Science, and Or Zuk from the Broad Institute of MIT and Harvard.

Host: Tomer Shlomi . ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400011200 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101117T140000 DTEND;TZID="Asia/Jerusalem":20101117T160000 DTSTAMP;TZID="Asia/Jerusalem":20101117T140000 FREEBUSY;FBTYPE=BUSY:20101117T140000/20101117T160000 SUMMARY;LANGUAGE=en-US:The msc talk by Dmitry Zinenko about Communication- Efficient Self-Stabilization Using Gossip. at 2010-11-17 14:00:00 DESCRIPTION;LANGUAGE=en-US:A self-stabilizing algorithm is a distributed a lgorithm that converges to a legal solution from any initial configuration . Most self-stabilizing protocols rely on checking every neighbor of ever y processor continuously to detect inconsitencies. Such protocols have a high communication cost, especially in dense networks. We investigate the potential usefulness of gossip for improving the communication efficie ncy of self-stabilizing protocols. We present randomized low communication self-stabilizing algorithms for several major tasks, namely, spanning tre e construction, distributed reset, and unison. The talk is self contained and requires basic knowledge of probability and distributed algorithms. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400011250 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101122T183000 DTEND;TZID="Asia/Jerusalem":20101122T203000 DTSTAMP;TZID="Asia/Jerusalem":20101122T183000 FREEBUSY;FBTYPE=BUSY:20101122T183000/20101122T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Boaz Goldstein abou t Haifux, Haifa Linux Club: The Cairo Graphics Compositing Library at 2010 -11-22 18:30:00 DESCRIPTION;LANGUAGE=en-US:The Cairo graphics library has become an integr al part of open source graphics, being the bases for quality graphics i n various projects, such as Gnome and Fire fox. This lecture will cover its abilities, future direction and give a brief intro to using it. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400011330 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101123T143000 DTEND;TZID="Asia/Jerusalem":20101123T163000 DTSTAMP;TZID="Asia/Jerusalem":20101123T143000 FREEBUSY;FBTYPE=BUSY:20101123T143000/20101123T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Tamir Hazan about Approximated L earning and Inference in Large Scale Graphical Models at 2010-11-23 14:30: 00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011300 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101124T122000 DTEND;TZID="Asia/Jerusalem":20101124T142000 DTSTAMP;TZID="Asia/Jerusalem":20101124T122000 FREEBUSY;FBTYPE=BUSY:20101124T122000/20101124T142000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Maurice Herlihy, Brown U niversity about Theory Seminar: The Topology of Shared-memory Adversaries at 2010-11-24 12:20:00 DESCRIPTION;LANGUAGE=en-US:Failure patterns in modern parallel and distrib uted system are not necessarily uniform. The notion of an adversary sche duler is a natural way to extend the classical wait-free and t-faulty mo dels of computation. A well-established way to characterize an adversary is by its set of cores, where a core is any minimal set of processes th at cannot all fail in any execution. We show that the protocol complex associated with an adversary is (c-2)-connected, where c is the size of the adversary's smallest core. This implies, among other results, that such an adversary can solve c-set agreement, but not (c-1)-set agreement . The proofs are combinatorial, relying on a novel application of the Ne rve Theorem of modern combinatorial topology.

Joint with Serg io Rajsbaum ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011320 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101124T140000 DTEND;TZID="Asia/Jerusalem":20101124T160000 DTSTAMP;TZID="Asia/Jerusalem":20101124T140000 FREEBUSY;FBTYPE=BUSY:20101124T140000/20101124T160000 SUMMARY;LANGUAGE=en-US:The phd talk by Rafi Chen about New Techniques for the Cryptanalysis of Hash Functions at 2010-11-24 14:00:00 DESCRIPTION;LANGUAGE=en-US:Cryptographic hash functions take a message of arbitrary length and generate a short fingerprint. Their main use are for digital signatures, due to their collision resistance property, i.e., that it is hard to find two different messages that have the same fingerprint. In this talk we present novel cryptanalysis techniques that we develop ed to attack hash functions. These techniques improve the functionality of differential attacks against the collision resistance property of the has h functions. We will briefly present the "neutral-bits" technique that ena bles an attacker to generate many messages that partially conform to a cha racteristic, and the "multi-block" technique that instead of directly atta cking a single block for a collision, creates a path of near-collisions th at can be found much more efficiently, and end with a collision. A major p art of the talk is dedicated to the introduction of our recent work on the "second-order differential" technique. Our techniques are generic and we confirmed them by constructing attacks on SHA-0 and SHA-1. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400011340 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101128T110000 DTEND;TZID="Asia/Jerusalem":20101128T130000 DTSTAMP;TZID="Asia/Jerusalem":20101128T110000 FREEBUSY;FBTYPE=BUSY:20101128T110000/20101128T130000 SUMMARY;LANGUAGE=en-US:The colloq talk by Dafna Shahaf SPECIAL GUEST LECTU RE about Connecting the Dots Between News Articles at 2010-11-28 11:00:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub NOTE UNUSUAL DAY AND TIME Bld. UID:111ec4410400011270 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101130T113000 DTEND;TZID="Asia/Jerusalem":20101130T133000 DTSTAMP;TZID="Asia/Jerusalem":20101130T113000 FREEBUSY;FBTYPE=BUSY:20101130T113000/20101130T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Jeremy Kaminski (CS, Holon I nstitute of Technology) about Pixel Club Seminar: Single Image Head Orient ation and Gaze Detection at 2010-11-30 11:30:00 DESCRIPTION;LANGUAGE=en-US:We introduce a system to compute both head orie ntation and gaze detection from a single image. The system uses a camera with fixed parameters and requires no user calibration.

Our approach to head orientation is based on a geometrical model of the huma n face, and is derived from morphological and physiological data. Eye ga ze detection is based on a geometrical model of the human eye. Two new a lgorithms are introduced that require either two or three feature points to be extracted from each image.

Our algorithms are robust a nd run in real-time on a typical PC, which makes our system useful for a large variety of needs, from driver attention monitoring to machine-hum an interaction. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011350 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101130T143000 DTEND;TZID="Asia/Jerusalem":20101130T163000 DTSTAMP;TZID="Asia/Jerusalem":20101130T143000 FREEBUSY;FBTYPE=BUSY:20101130T143000/20101130T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Maurice Herlihy about Applicatio ns of Shellable Complexes to Distributed Computing at 2010-11-30 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011390 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101206T183000 DTEND;TZID="Asia/Jerusalem":20101206T203000 DTSTAMP;TZID="Asia/Jerusalem":20101206T183000 FREEBUSY;FBTYPE=BUSY:20101206T183000/20101206T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Dalit Ken-Dror (Hai fa University) about Haifux, Haifa Linux Club: Creative Commons Licenses, Open Source Software Licenses & the new Israeli Copyright Act of 2007 at 2010-12-06 18:30:00 DESCRIPTION;LANGUAGE=en-US:TBA ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400011490 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101207T113000 DTEND;TZID="Asia/Jerusalem":20101207T133000 DTSTAMP;TZID="Asia/Jerusalem":20101207T113000 FREEBUSY;FBTYPE=BUSY:20101207T113000/20101207T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Eyal Madar (EE, Technion) ab out Pixel Club Seminar: Combined Local-Global Background Modeling for Anom aly Detection in Hyperspectral Images at 2010-12-07 11:30:00 DESCRIPTION;LANGUAGE=en-US:In this research, we address the problem of ano maly detection using remotely sensed spectral information collected by hyperspectral sensors. Anomaly detection algorithms first model the abu ndant material spectra (background). Then, every pixel spectrally differen t in a meaningful way from the background is declared to be an anomaly pixel. Two major approaches to statistical background modeling can be d istinguished: “the local approach” and “the global approach”. Loc al algorithms can tightly fit the background process but are subject to an over-fitting problem, which may produce an excessive number of false-alar ms. Global methods are more resistant to over-fitting, however, they have a limited ability to adapt to all nuances of the background process (an un der-fitting problem), which may result in high false alarm rates, as well as low probability of detection.

In our work, we propose a co mbination of the local and global background modeling approaches by introd ucing the BEVA (Background Extreme Value Analysis) algorithm. In its local part, the background process is estimated using a greedy sequential me thod. It is composed of robust estimation of the Gaussian statistics an d a background cluster hypothesis discriminator, based on Extreme Value Theory results. In its global part, the obtained local background models are inter-related to reduce the number of false alarms. In addition, we improve BEVA's local part via a preprocessing segmentation that is bas ed on Spectral Clustering. We also introduce the NG-BEVA algorithm; a n on-Gaussian version of BEVA that combines Extreme Value Theory results wit h Gamma distribution fitting. NG-BEVA is found to improve further the performance. 2010-12-06 18:30:00

* M.Sc. Research under th e supervision of Prof. David Malah and Dr. Meir Barzohar ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400011410 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101208T123000 DTEND;TZID="Asia/Jerusalem":20101208T143000 DTSTAMP;TZID="Asia/Jerusalem":20101208T123000 FREEBUSY;FBTYPE=BUSY:20101208T123000/20101208T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Jakob Nordstrom (KTH Roy al Institute of Technology) about Theory Seminar: On the Semantics of Loca l Characterizations for Linear-Invariant Properties at 2010-12-08 12:30:0 0 DESCRIPTION;LANGUAGE=en-US:A property of functions on a vector space is sa id to be linear-invariant if it is closed under linear transformations of the domain. Linear-invariant properties are some of the most well-studi ed properties in the field of property testing. Testable linear-invari ant properties can always be characterized by so-called local constraints, and of late there has been a rapidly developing body of research inves tigating the testability of linear-invariant properties in terms of the ir descriptions using such local constraints. One problematic aspect that has been l argely ignored in this line of research, however, is that syn tactically distinct local characterizations need not at all correspond to semantically distinct properties.

In fact, there are know n fairly dramatic examples where seemingly infinite families of properties collapse into a small finite set that was already well-understood. In this work, we therefore initiate a systematic study of the semantics of local characterizations of linear-invariant properties. For such properti es the local characterizations have an especially nice structure in ter ms of forbidden patterns on linearly dependent sets of vectors, which c an be encoded formally as matroid constraints. We develop techniques for determining, given two such matroid constraints, whether these constrain ts encode identical or distinct properties, and show for a fairly broad class of properties that these techniques provide necessary and suffic ient conditions for deciding between the two cases. We use these tools to show that recent ( yntactic) testability results indeed provide an infi nite number of infinite strict hierarchies of )semantically) distinct t estable locally characterized linear-invariant properties.

Jo int work with Arnab Bhattacharyya, Elena Grigorescu, and Ning Xie ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011430 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101214T113000 DTEND;TZID="Asia/Jerusalem":20101214T133000 DTSTAMP;TZID="Asia/Jerusalem":20101214T113000 FREEBUSY;FBTYPE=BUSY:20101214T113000/20101214T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Ora Gendler (EE, Technion) a bout Pixel Club Seminar: Toward Optimal Real-Time Transcoding of Images an d Video Data Using Requantization at 2010-12-14 11:30:00 DESCRIPTION;LANGUAGE=en-US:Visual communication often requires adaptation of the transmission bit-rate to the available channel bandwidth or display characteristics of the end users. In this work, requantization for transr ating of MPEG video sequences and JPEG still images is analyzed. We show t hat both the rate and the distortion of requantized images and video depen d mainly on the ratio between the new and the original quantization steps. Our analysis is based on the structure of the quantizer and the Laplace -like distribution of the DCT coefficients in sub-band coding. We also sho w that the rounding policy has a significant effect on the requantization performance. In addition, we examine a transrating approach that incorpora tes image downscaling along with requantization. This method is based on t he well known coding trade-off between the number of quantization levels a nd the number of pixels used to code an image. We propose a method that ef ficiently selects one of the coding modes: requantization or downscaling . We demonstrate how to incorporate the proposed requantization approach i nto a rate-control system using the Rho model as an example. We show that the resulting bit-rate of the proposed requantization method can be accu rately predicted using this model. Our conclusion is that the new approach could be instrumental in achieving a required bit-rate when recompressing still images and video at low distortion, while allowing real-time implem entation due to its low computational complexity.

M.Sc Thesis under the supervision of Dr. Moshe Porat. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400011420 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101214T123000 DTEND;TZID="Asia/Jerusalem":20101214T143000 DTSTAMP;TZID="Asia/Jerusalem":20101214T123000 FREEBUSY;FBTYPE=BUSY:20101214T123000/20101214T143000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Alan Winfield (Hewlett-Packa rd Professor of Electronic Engineering and Director of the Science Comm unication Unit at the University of the West) about Pixel Club Seminar: Topics in Distributed Robotics at 2010-12-14 12:30:00 DESCRIPTION;LANGUAGE=en-US:TBD ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 401 UID:111ec4410400011570 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101214T143000 DTEND;TZID="Asia/Jerusalem":20101214T163000 DTSTAMP;TZID="Asia/Jerusalem":20101214T143000 FREEBUSY;FBTYPE=BUSY:20101214T143000/20101214T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Liuba Shrira about Retro: An "Al way There" and "Not in The Way" Snapshot System at 2010-12-14 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011480 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101215T113000 DTEND;TZID="Asia/Jerusalem":20101215T133000 DTSTAMP;TZID="Asia/Jerusalem":20101215T113000 FREEBUSY;FBTYPE=BUSY:20101215T113000/20101215T133000 SUMMARY;LANGUAGE=en-US:The cggc talk by Yaron Lipman (Princeton University ) about CGGC Seminar: Surface Comparison using Conformal Geometry at 2010- 12-15 11:30:00 DESCRIPTION;LANGUAGE=en-US:In this talk we will present three applications of conformal geometry to the problems of surface matching and compariso n. In particular we will show how certain ideas originated in the theory of conformal geometry can be used to define novel metrics measuring dis similarities and finding correspondences betweens pairs of surfaces auto matically.

The prominent rigidity of conformal mappings rend ers these metrics computationally efficient, add suggests that conformal geometry can play a role in reducing the complexity of certain surface matching paradigms. We report results on few datasets of biological surf aces as well as other common 3D surface datasets. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 5 (Floor 1) UID:111ec4410400011500 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101215T122000 DTEND;TZID="Asia/Jerusalem":20101215T142000 DTSTAMP;TZID="Asia/Jerusalem":20101215T122000 FREEBUSY;FBTYPE=BUSY:20101215T122000/20101215T142000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Klim Efremenko (Tel-Aviv University) about Theory Seminar: Local List-Decoding with a Constant Nu mber of Queries at 2010-12-15 12:20:00 DESCRIPTION;LANGUAGE=en-US:Recently there was constructed locally-decoda ble codes of sub-exponential length. This result showed that these codes c an handle up to one third fraction of errors. In this talk we show that th e same codes can be locally unique-decoded from error rate upto half and l ocally list-decoded from error rate $1-\alpha$ for any $\alpha>0$, with on ly a constant number of queries and a constant alphabet size. This gives t he first sub-exponential codes that can be locally list-decoded with a con stant number of queries. We also will show a generic, simple way to amplif y the error-tolerance of any locally decodable code. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011440 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101215T133000 DTEND;TZID="Asia/Jerusalem":20101215T153000 DTSTAMP;TZID="Asia/Jerusalem":20101215T133000 FREEBUSY;FBTYPE=BUSY:20101215T133000/20101215T153000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Eric Price (MIT) about Pixel Club Seminar: Efficient Linear Sketches for the Set Query Problem at 2010 -12-15 13:30:00 DESCRIPTION;LANGUAGE=en-US:Sparse recovery, or compressive sensing, is the problem of approximating a vector x from a low-dimensional linear sket ch Ax, where A is an m x n matrix with m << n. We cannot recover x exa ctly, so we require that, if x is "close" to being "good", the recovere d vector is "close" to x. Essentially optimal solutions are known when "close" uses l_1 or l_2 distance and "good" means being k-sparse, but no t for other definitions of "close" and "good".

We give a f ramework for sparse recovery where the problem is split into two parts. In the first part we locate the large coefficients of x, and in the s econd part we estimate their values (termed the "set query problem"). We give an optimal algorithm for the set query problem, leaving only th e first part for problem-specific development. We then give improved a lgorithms for locating the large coefficients when x is close to being (1) distributed according to a power law distribution or (2) block spar se. In contrast to methods based on pseudoinverses or dense matrices, our constructions use sparse matrices and support linear or nearly line ar recovery time. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400011530 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101215T163000 DTEND;TZID="Asia/Jerusalem":20101215T183000 DTSTAMP;TZID="Asia/Jerusalem":20101215T163000 FREEBUSY;FBTYPE=BUSY:20101215T163000/20101215T183000 SUMMARY;LANGUAGE=en-US:The msc talk by Ariel Kulik about Approximations fo r Monotone and Non-monotone Submodular Maximization with Knapsack Constrai nts at 2010-12-15 16:30:00 DESCRIPTION;LANGUAGE=en-US:Submodular maximization generalizes many fundam ental problems in discrete optimization, including Max-Cut in directed/u ndirected graphs, maximum coverage, maximum facility location and market ing over social networks. We consider the problem of maximizing any submodular function subject to $d$ knapsack constraints, where $d$ is a fixed constant. We establish a strong relation between the discrete p roblem and its continuous relaxation, obtained through {\em extension by expectation} of the submodular function. Formally, we show that, for an y non-negative submodular function, an $\alpha$-approximation algorithm for the continuous relaxation implies a randomized $(\alpha - \eps)$-app roximation algorithm for the discrete problem. We use this relation to i mprove the best known approximation ratio for the problem to $1/4- \eps$ , for any $\eps > 0$, and to obtain a nearly optimal $(1-e^{-1}-\eps)- $approximation ratio for the monotone case, for any $\eps>0$. We fu rther show that the probabilistic domain defined by a continuous solutio n can be reduced to yield a polynomial size domain, given an oracle for the extension by expectation. This leads to a deterministic version of o ur technique. Our approach has a potential of wider applicability, wh ich we demonstrate on the examples of the Generalized Assignment Problem and Maximum Coverage with additional knapsack constraints. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec441040007250 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101216T113000 DTEND;TZID="Asia/Jerusalem":20101216T133000 DTSTAMP;TZID="Asia/Jerusalem":20101216T113000 FREEBUSY;FBTYPE=BUSY:20101216T113000/20101216T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Steve Seitz (University of W ashington) about Pixel Club Seminar: Reconstructing the World from Photos on the Internet at 2010-12-16 11:30:00 DESCRIPTION;LANGUAGE=en-US:‏There's a big difference between looking at a photograph of a place and being there. But what if you had access to a d atabase of every possible image of that place and could conjure up any vie w at will? With billions of photographs currently available online, the In ternet is beginning to resemble such a database, capturing our world's sit es from a huge number of vantage points and viewing conditions. For exampl e, a Google image search for "notre dame" or "grand canyon" each return mi llions of photos, showing the sites from myriad viewpoints, different time s of day and night, and changes in season, weather and decade.This talk ex plores ways of transforming this massive, unorganized photo collection int o 3D scene reconstructions and visualizations of the world's sites, cities , and landscapes. After a brief recap of our work on Photo Tourism and Pho tosynth, I will focus on current efforts and newest results in the domains of city-scale 3D reconstruction and new visual interfaces for navigating photo collections. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400011540 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101220T183000 DTEND;TZID="Asia/Jerusalem":20101220T203000 DTSTAMP;TZID="Asia/Jerusalem":20101220T183000 FREEBUSY;FBTYPE=BUSY:20101220T183000/20101220T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Dalit Ken-Dror (Hai fa University) about Haifux, Haifa Linux Club: Creative Commons Licenses, Open Source Software Licenses & the new Israeli Copyright Act of 2007 at 2 010-12-20 18:30:00 DESCRIPTION;LANGUAGE=en-US:Part I of the talk covered the new Israeli Copy right Act of 2007. This part of the talk will focus on Creative Commons Li censes. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400011610 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101222T123000 DTEND;TZID="Asia/Jerusalem":20101222T143000 DTSTAMP;TZID="Asia/Jerusalem":20101222T123000 FREEBUSY;FBTYPE=BUSY:20101222T123000/20101222T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Nir Ailon (CS, Technion) about Theory Seminar: Almost Optimal Unrestricted Fast Johnson-Lindenstra uss Transform at 2010-12-22 12:30:00 DESCRIPTION;LANGUAGE=en-US:The problems of random projections and sparse r econstruction have much in common and individually received much attenti on. Surprisingly, until now they progressed in parallel and remained mos tly separate. Here, we employ new tools from probability in Banach space s that were successfully used in the context of sparse reconstruction to advance on an open problem in random pojections. In particular, we gene ralize and use an intricate result by Rudelson and Vershynin for sparse reconstruction which uses Dudley’s theorem for bounding Gaussian proce sses. Our main result states that any set of N = exp(O(n)) real vectors in n dimensional space can be linearly mapped to a space of dimension k = O(logN polylog(n)), while (1) preserving the pairwise distances among the vectors to within any constant distortion and (2) being able to appl y the transformation in time O(n log n) on each vector. This improves on the best known N = exp(O(n^{1/2})) achieved by Ailon and Liberty and N = exp(O(n^{1/3})) by Ailon and Chazelle. The dependence in the distortion constant however is believed to be suboptimal and subject to further i nvestigation. For constant distortion, this settles the open question pose d by these authors up to a polylog(n) factor while considerably simplify ing their construction.

Joint work with Edo Liberty. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011560 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101222T140000 DTEND;TZID="Asia/Jerusalem":20101222T160000 DTSTAMP;TZID="Asia/Jerusalem":20101222T140000 FREEBUSY;FBTYPE=BUSY:20101222T140000/20101222T160000 SUMMARY;LANGUAGE=en-US:The msc talk by Gleb Polevoy about Bandwidth Alloca tion in Cellular Networks with Multiple Interferences at 2010-12-22 14:00: 00 DESCRIPTION;LANGUAGE=en-US:We study the problem of bandwidth allocation wi th multiple interferences. In this problem the input consists of a set o f users and a set of base stations. Each user has a list of requests, ea ch consisting of a base station, a frequency demand, and a profit that m ay be gained by scheduling this request. The goal is to find a maximum p rofit set of user requests $\calS$ that satisfies the following condition s: 1) $\calS$ contains at most one request per user, 2) the frequency sets allotted to requests in $\calS$ that correspond to the same b ase station are pairwise non-intersecting, and 3) the QoS received by any user at any frequency is reasonable according to an interference m odel. We show that these problems are extremely hard to approximate if the interferences depend on both the interfered and the interfering base stations. On the other hand, we provide constant factor approximation a lgorithms for the case where the interferences depend only on the interfe ring base stations. We employ the local ration technique. We also consi der a restrictive special case that is closely related to the knapsack pr oblem. We show that this special case is NP-hard and that it admits an F PTAS. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400011550 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101222T143000 DTEND;TZID="Asia/Jerusalem":20101222T163000 DTSTAMP;TZID="Asia/Jerusalem":20101222T143000 FREEBUSY;FBTYPE=BUSY:20101222T143000/20101222T163000 SUMMARY;LANGUAGE=en-US:The CSpecial Talk talk by Prof. Eli Gafni (UCLA) ab out CSpecial Talk: Topology in Distributed Computing at 2010-12-22 14:30:0 0 DESCRIPTION;LANGUAGE=en-US:The integer 1 is the unity of multiplication. T he integer 12 represents the senior high-school year. They surely are diff erent in many respects. But distributed computing does not multiply, nei ther it attends school, so why should it distinguish between the two int egers? I refer to the nice generalization of consensus introduced by Cha udhuri called k-set consensus. While in consensus processors vote for a un ique participating processor to be their representative, in set consensus their vote may diverge but at most to k will be voted for. Are there pro perties that hold for consensus but do not have a corresponding property f or k-set consensus? This talk will survey a slew of interesting results ar ising from the conviction that for distributed computing 1 is not differen t than 12.

The talk will survey results with: Afek, Travers, Guerraoui, Kuznetsov, Raynal, and Rajsbaum.

Bio: Eli Gafni re ceived his first degree form the Technion, second from UIUC, and third fro m MIT, all in E.E. He was involved with the Internet in the early days whe n it consisted of only few nodes. Unlike his contemporaries in MIT, of whi ch quite a few went on to become few hundred times Internet Millionaires, he joined UCLA computer-science department and abstracted the Internet to the point that he became even too theoretical for that discipline. Neverth eless, with tenure, he is still a Professor at UCLA, holding forth that in tellectual fun or the ability to roam perhaps aimlessly through intellectu ally challenging roads, is the reason to be in University rather than Indu stry. He does not envy the Millionaires, and he is only partially responsi ble for the sorry financial state of the UC system. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 5 UID:111ec4410400011630 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101223T143000 DTEND;TZID="Asia/Jerusalem":20101223T163000 DTSTAMP;TZID="Asia/Jerusalem":20101223T143000 FREEBUSY;FBTYPE=BUSY:20101223T143000/20101223T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Ariel Procaccia about Pixel Club Seminar: May the Optimal Candidate Win! Optimization under Social Choice Constraints at 2010-12-23 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011400 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101226T123000 DTEND;TZID="Asia/Jerusalem":20101226T143000 DTSTAMP;TZID="Asia/Jerusalem":20101226T123000 FREEBUSY;FBTYPE=BUSY:20101226T123000/20101226T143000 SUMMARY;LANGUAGE=en-US:The colloq talk by Daniel M. Berry SPECIAL LECTURE note unusual hour about Requirements Determination is Unstoppable: An Expe rience Report at 2010-12-26 12:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011510 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101226T143000 DTEND;TZID="Asia/Jerusalem":20101226T163000 DTSTAMP;TZID="Asia/Jerusalem":20101226T143000 FREEBUSY;FBTYPE=BUSY:20101226T143000/20101226T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Dan Kushnir about Anisotropic di ffusion maps of sub-manifolds with applications at 2010-12-26 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011360 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101228T103000 DTEND;TZID="Asia/Jerusalem":20101228T123000 DTSTAMP;TZID="Asia/Jerusalem":20101228T103000 FREEBUSY;FBTYPE=BUSY:20101228T103000/20101228T123000 SUMMARY;LANGUAGE=en-US:The cggc talk by Mirela Ben-Chen (Stanford Universi ty) about CGGC Seminar: AKVFs - A New Computational Tool for Approximate S hape at 2010-12-28 10:30:00 DESCRIPTION;LANGUAGE=en-US:A fundamental problem in geometry processing is to characterize the possible transformations a shape can undergo, and s till remain "the same". Such transformations are called isometries, sinc e they preserve some notion of distance between all points on the shape. In many cases, however, especially when dealing with discrete shapes, approximate isometries better characterize the wealth of shapes which re present the same object, such as different poses of a character. Such ma ps are harder to represent and parameterize, in particular when consider ing geodesic distances - along paths on the shape.

We tack le this challenge by using Killing vector fields - a special type of vec tor fields which are the generators of isometries. We show how approxima te continuous isometries can be parameterized by approximate Killing vec tor fields (AKVFs), and how to find those efficiently on a discrete shap e. Our approach gives rise to a new operator, whose spectral data convey s much information about a shape. We show how these insights on represen ting shape isometries can be applied to various problems in geometry pro cessing, such as texture generation, segmentation and deformation. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 7 UID:111ec4410400011620 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101228T103000 DTEND;TZID="Asia/Jerusalem":20101228T123000 DTSTAMP;TZID="Asia/Jerusalem":20101228T103000 FREEBUSY;FBTYPE=BUSY:20101228T103000/20101228T123000 SUMMARY;LANGUAGE=en-US:The colloq talk by Mirela Ben-Chen about AKVFs - A New Computational Tool for Approximate Shape Isometries at 2010-12-28 10:3 0:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room Taub 7 Taub Bld. UID:111ec4410400011600 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101228T113000 DTEND;TZID="Asia/Jerusalem":20101228T133000 DTSTAMP;TZID="Asia/Jerusalem":20101228T113000 FREEBUSY;FBTYPE=BUSY:20101228T113000/20101228T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Hod Lipson (Mechanical & Aer ospace Engineering and Computing & Information Science at Cornell Universi ty) about Pixel Club: Voxel Fabbing: From Analog to Digital 3D Printing at 2010-12-28 11:30:00 DESCRIPTION;LANGUAGE=en-US:The transition from analog to digital has revol utionized many fields over the past century – most notably computation a nd communication – and can be used to similarly revolutionize additive m anufacturing technology. In contrast with continuous (analog) materials pr oduced by traditional rapid prototyping, digital materials are composed of many discrete, self-aligning voxels placed in a massively parallel layer deposition process. Digital principles allow for perfect replication and z ero noise despite using a noisy and inaccurate substrate. The paradigm of digital printing with prefabricated voxels enables parts composed of multi ple materials with mutually incompatible processing characteristics and sp ecific functionality to be combined in a single freeform fabrication proce ss. This talk will describe the concept as well as demonstrated a proof of concept digital fabricator capable of printing spheres of multiple materi als at mm-scale resolution, including printing of hybrid metal/polymer par ts. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011700 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101228T143000 DTEND;TZID="Asia/Jerusalem":20101228T163000 DTSTAMP;TZID="Asia/Jerusalem":20101228T143000 FREEBUSY;FBTYPE=BUSY:20101228T143000/20101228T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Danny Bickson about GraphLab: As ynchronous Graph Computation in the Clouds and Beyond at 2010-12-28 14:30: 00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011370 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101228T150000 DTEND;TZID="Asia/Jerusalem":20101228T170000 DTSTAMP;TZID="Asia/Jerusalem":20101228T150000 FREEBUSY;FBTYPE=BUSY:20101228T150000/20101228T170000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Doron Lipson (Foun dation Medicine Inc.) about Bioinformatics Forum (NOTE UNUSUAL HOUR): Sing le Molecule Sequencing Based Digital Gene Expression - Investigating the Transcriptome through a Clearer Lens at 2010-12-28 15:00:00 DESCRIPTION;LANGUAGE=en-US:Digital Gene Expression (DGE) profiling by next -generation sequencing (NGS) technologies has the potential of becoming the leading method for quantitative analysis of complete transcriptomes. While similar approaches have been implemented in the past by methods s uch as SAGE, one of the key advantages of NGS technologies is an extreme ly high throughput which enables accurate quantification over a wide dyn amic range of transcript abundances, as well as relatively low sample pr eparation bias.

In this talk I will describe the molecular a nd computational aspects of a single-molecule sequencing based approach to DGE and its application to model yeast and human transcriptomes. In f ollowing studies, the ability of this method to exhaustively characteriz e the transcriptome has led to the discovery of a novel type of gene-ter mini-associated small RNA as well as a potential new bio-marker for earl y detection of pancreatic cancer.

Host: Dr. Zohar Yakhini. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400011590 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101229T113000 DTEND;TZID="Asia/Jerusalem":20101229T133000 DTSTAMP;TZID="Asia/Jerusalem":20101229T113000 FREEBUSY;FBTYPE=BUSY:20101229T113000/20101229T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Gilad Lerman (School of Math ematics, University of Minnesota) about Pixel Club Seminar: Multi-Manif old Data Modeling: Foundations and Applications at 2010-12-29 11:30:00 DESCRIPTION;LANGUAGE=en-US:We present several methods for multi-manifold d ata modeling, i.e., modeling data by mixtures of possibly intersecting m anifolds. We focus on algorithms for the special case where the underlyi ng manifolds are affine or linear subspaces. We emphasize various theo retical results supporting the performance of some of these algorithms, in particular their robustness to noise and outliers. We demonstrate how such theoretical insights guide us in practical choices and present app lications of such algorithms.

This is part of joint works wit h E. Arias-Castro, S. Atev, G. Chen, A. Szlam, Y. Wang, T. Whitehouse an d T. Zhang ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011580 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101229T122000 DTEND;TZID="Asia/Jerusalem":20101229T142000 DTSTAMP;TZID="Asia/Jerusalem":20101229T122000 FREEBUSY;FBTYPE=BUSY:20101229T122000/20101229T142000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Noam Livne (Weizmann Ins titute of Science) about Theory Seminar: Sequential Rationality in Cryptog raphic Protocols at 2010-12-29 12:20:00 DESCRIPTION;LANGUAGE=en-US:Much ofMuch of the literature on rational crypt ography focuses on analyzing the strategic properties of cryptographic p rotocols. However, due to the presence of computationally-bounded player s and the asymptotic nature of cryptographic security, a definition of s equential rationality for this setting has thus far eluded researchers.

We propose a new framework for overcoming these obstacles, an d provide the first definitions of computational solution concepts that guarantee sequential rationality. We argue that natural computational va riants of subgame perfection are too strong for cryptographic protocols. As an alternative, we introduce a weakening called threat-free Nash equ ilibrium that is more permissive but still eliminates the undesirable `` empty threats'' of non-sequential solution concepts.

To dem onstrate the applicability of our framework, we revisit the problem of i mplementing a mediator for correlated equilibria (Dodis-Halevi-Rabin, Cr ypto'00), and propose a variant of their protocol that is sequentially r ational for a non-trivial class of correlated equilibria. Our treatment provides a better understanding of the conditions under which mediators in a correlated equilibrium can be replaced by a stable protocol. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011470 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101230T113000 DTEND;TZID="Asia/Jerusalem":20101230T133000 DTSTAMP;TZID="Asia/Jerusalem":20101230T113000 FREEBUSY;FBTYPE=BUSY:20101230T113000/20101230T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Michael Lustig (Electrical E ngineering and Computer Sciences, College of Engineering, UC Berkeley) about Pixel Club Seminar: Practical Parallel Imaging Compresses Sens ing MRI: Summary of Two Years of Experience Accelerating Body MRI of Pedia tric Patients at 2010-12-30 11:30:00 DESCRIPTION;LANGUAGE=en-US:Magnetic Resonance Imaging has revolutionized d iagnostic medicine. It is an excellent tool for disease diagnosis and moni toring, offering superb soft tissue contrast and high anatomic resolution; unlike computed tomography (CT), it lacks of ionizing radiation. However MRI suffers from several shortcomings, one of which is the inherently slow data acquisition. This has limited the penetration of MRI to applications that require sharp images of fast moving small body parts, such as angiog raphy, cardiovascular imaging, imaging small children and fetal imaging.

To overcome this limitation, many methods for fast imaging by reduced sampling have been proposed. These are based on exploiting various redundancies in the data. Parallel Imaging is the most noteworthy. It is a well-established accelerated imaging technique based on the spatial sen sitivity of array receivers. Compressed sensing is an emerging accelerated imaging based on the compressibility of medical images. Synergistic combi nation of parallel imaging and compressed sensing offers much higher accel eration and better quality imagery.

For the last two years, w e have been experimenting with applying compressed sensing parallel imagin g for MR body imaging of pediatric patients. It is a joint-effort by teams from UC Berkeley, Stanford University and GE Healthcare. The talk aims to summarize our experience so far. I will start with some background on MR imaging, parallel imaging and compressed sensing MRI. I will then turn to describe our unique approach of data acquisition and reconstruction, our implementation on parallel processors (multi-core and GPGPU), applications and clinical studies. Our approach is implemented and installed at Lucile Packard Children's Hospital at Stanford. So far, it is the first and only place in the world in which compressed sensing is routinely used in clini cal practice. We can routinely achieve clinical quality reconstructions of body pediatric MR, at more than 8-fold acceleration. These are reconstruc ted and displayed in about a minute using our GPU-based parallel processin g reconstruction.

Time permitting I will also describe some r ecent advances in parallel imaging that leads to interesting low-rank stru ctured matrix completion problems. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011710 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20101230T143000 DTEND;TZID="Asia/Jerusalem":20101230T163000 DTSTAMP;TZID="Asia/Jerusalem":20101230T143000 FREEBUSY;FBTYPE=BUSY:20101230T143000/20101230T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Ishai Menache about Efficient cl oud operations via job-migration and pricing at 2010-12-30 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011720 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110102T143000 DTEND;TZID="Asia/Jerusalem":20110102T163000 DTSTAMP;TZID="Asia/Jerusalem":20110102T143000 FREEBUSY;FBTYPE=BUSY:20110102T143000/20110102T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Yoram Bachrach about Conspiracie s, Cooperation and Power at 2011-01-02 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011730 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110103T183000 DTEND;TZID="Asia/Jerusalem":20110103T203000 DTSTAMP;TZID="Asia/Jerusalem":20110103T183000 FREEBUSY;FBTYPE=BUSY:20110103T183000/20110103T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Raz Ben Yehuda abo ut Haifux Club: From VxWorks To Linux at 2011-01-03 18:30:00 DESCRIPTION;LANGUAGE=en-US:Preempt RT is a growing hard real time linux OS . In this session I will present Preempt RT basics and uniqueness, provide a Live demo, present a benchmark of a company that is migrating from vxwo rks to Preempt RT and discuss the non-techinical aspects of migrating to l inux. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400011690 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110104T113000 DTEND;TZID="Asia/Jerusalem":20110104T133000 DTSTAMP;TZID="Asia/Jerusalem":20110104T113000 FREEBUSY;FBTYPE=BUSY:20110104T113000/20110104T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Benjamin Kimia (Brown Univer sity) about Pixel Club: Perceptual Fragments: Bottom-Up and Top-Down Use o f Shape in Object Recognition at 2011-01-04 11:30:00 DESCRIPTION;LANGUAGE=en-US:The bottom-up “segmentation followed by recog nition” strategy has for some time now given way to feature-based discri minative recognition with significant success. As the number of categories and exemplars per category increases, however, low-level features are no longer sufficiently discriminative, motivating the construction and use of more complex features. It is argued here that these complex features will necessarily be encoding shape and this in turn requires curves and region s, thus reviving aspects of bottom-up segmentation strategies. We suggest that the demise of segmentation was due to prematurely committing to a gro uping in face of ambiguities and propose a framework for representing mu ltiple grouping options in a containment graph. Specifically, we use conto ur symmetry to partition the image into atomic fragments and define transf orms to iteratively grow these atomic fragments into mote distinctive perc eptual fragments, the nodes of the containment graph. We also briefly pres ent a fragment-based language for generating shapes and the use of fragmen ts in top-down category recognition. The bottom-up and top-down processes are then integrated by interaction through the mid-level representation of perceptual fragments. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012280 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110104T123000 DTEND;TZID="Asia/Jerusalem":20110104T143000 DTSTAMP;TZID="Asia/Jerusalem":20110104T123000 FREEBUSY;FBTYPE=BUSY:20110104T123000/20110104T143000 SUMMARY;LANGUAGE=en-US:The colloq talk by Scott Aaronson SPECIAL LECTURE a bout The Computational Complexity of Linear Optics at 2011-01-04 12:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 401 Taub NOTE UNUSUAL TIME AND PLACE Bld. UID:111ec4410400012270 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110104T143000 DTEND;TZID="Asia/Jerusalem":20110104T163000 DTSTAMP;TZID="Asia/Jerusalem":20110104T143000 FREEBUSY;FBTYPE=BUSY:20110104T143000/20110104T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Shahar Dobzinski about Building a Bridge between Incentives and Computation at 2011-01-04 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011520 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110105T132000 DTEND;TZID="Asia/Jerusalem":20110105T152000 DTSTAMP;TZID="Asia/Jerusalem":20110105T132000 FREEBUSY;FBTYPE=BUSY:20110105T132000/20110105T152000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Udi Weider (Microsoft Re search) about Theory Seminar: Lower Bounds on Near Neighbor Search via Met ric Expansion at 2011-01-05 13:20:00 DESCRIPTION;LANGUAGE=en-US:We show that the cell probe complexity of perfo rming nearest neighbor (NNS) search on a metric space is tightly related t o the expansion of the metric space: Given a metric space we look at the g raph obtained by connecting every pair of points within a certain distance r. We then look at various notions of expansion in this graph relating th em to the cell probe complexity of NNS for randomized and deterministic, e xact and approximate algorithms. These relationships can be used to derive most of the known lower bounds in the well known metric spaces such as l1 , l2, l1 by simply computing their expansion, as well as several new ones. Our work essentially reduces the problem of proving cell probe lower bou nds of near neighbor search to that of computing the appropriate expansion parameter.

Joint work with Rina Panigrahy and Kunal Talwar ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 (note unusual room) UID:111ec4410400011750 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110106T143000 DTEND;TZID="Asia/Jerusalem":20110106T163000 DTSTAMP;TZID="Asia/Jerusalem":20110106T143000 FREEBUSY;FBTYPE=BUSY:20110106T143000/20110106T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Noam Rinetzky SPECIAL LECTURE ab out Verifying Linearizability with Hindsight at 2011-01-06 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011670 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110109T113000 DTEND;TZID="Asia/Jerusalem":20110109T133000 DTSTAMP;TZID="Asia/Jerusalem":20110109T113000 FREEBUSY;FBTYPE=BUSY:20110109T113000/20110109T133000 SUMMARY;LANGUAGE=en-US:The phd talk by Eshcar Hilel about Concurrent Data Structures: Methodologies and Inherent Limitations at 2011-01-09 11:30:00 DESCRIPTION;LANGUAGE=en-US:As multi-core and multiprocessing architectures are becoming common, modern applications require concurrent data structu res for their computations. Designing concurrent data structures and ensu ring their correctness is a difficult task; significantly more challengin g than doing so for their sequential counterparts. Transactional memor y (TM), a programming model in which concurrent processes synchronize via in-memory transactions, is one prominent approach for alleviating the di fficulty of programming concurrent data structures for multi-core and mul tiprocessing systems. TM is seriously considered as part of software solu tions and as a basis for novel hardware designs. It is therefore imperati ve to understand inherent tradeoffs in the design and implementation of t ransactional memory. In the talk, we present two inherent limitations o n implementations of transactional memory. First, we discuss TMs avoiding interference between transactions with disjoint data. We prove that in these implementations, read-only transactions that always terminate succe ssfully, not only have to write to the memort, but the number of writes is linear in the number of items the transaction is reading. Allowing to access the same memory locations from inside and outside a transaction is crucial both for interoperability with legacy code and in order to im prove the performance of the TM. Supporting privatization, allows to iso late some items making them private to a process; the process can thereaf ter access them nontransactionally, without interference by other process es. We show privatization has an inherent cost, linear in the number of p rivatized items. The assumptions needed to prove the bounds indicate that limiting the parallelism of the TM or tracking the items of other transa ctions are the price to pay for efficient privatization. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 401 UID:111ec4410400012350 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110109T123000 DTEND;TZID="Asia/Jerusalem":20110109T143000 DTSTAMP;TZID="Asia/Jerusalem":20110109T123000 FREEBUSY;FBTYPE=BUSY:20110109T123000/20110109T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Raghu Meka (University o f Texas at Austin) about Theory Seminar: Pseudorandom Generators from Inva riance Principles at 2011-01-09 12:30:00 DESCRIPTION;LANGUAGE=en-US:Invariance principles or limit theorems have re cently found several important applications in theoretical computer sci ence. In this talk I'll present some recent results with the broad theme o f constructing pseudorandom generators from invariance principles. The first set of results concerns the natural question of constructing pseu dorandom generators (PRGs) for low-degree polynomial threshold functions ( PTFs). We give a PRG with seed-length log n/eps^{O(d)}$ fooling degre e d PTFs with error at most eps. Previously, no nontrivial constructions w ere known even for quadratic threshold functions and constant error eps. For the class of degree 1 threshold functions or halfspaces, we construct PRGs with much better dependence on the error parameter eps and obtain a PRG with seed-length O(log n + log^2(1/eps)). The second set of results co ncerns "discrete central limit theorems", and fooling linear forms over 0- 1 inputs in total variation distance.

Joint work with David Z uckerman; Parikshit Gopalan, Omer Reingold and David Zuckerman. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 539 UID:111ec4410400012310 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110111T113000 DTEND;TZID="Asia/Jerusalem":20110111T133000 DTSTAMP;TZID="Asia/Jerusalem":20110111T113000 FREEBUSY;FBTYPE=BUSY:20110111T113000/20110111T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Tammy Avraham (CS, Technion) about Pixel Club: Non-Local Characterization of Scenery Images: Statistic s, 3D Reasoning, and a Generative Model at 2011-01-11 11:30:00 DESCRIPTION;LANGUAGE=en-US:This work focuses on characterizing scenery ima ges. We semantically divide the objects in natural landscape scenes into b ackground and foreground and show that the shapes of the regions associate d with these two types are statistically different. We then focus on the b ackground regions. We study statistical properties such as size and shape, location and relative location, the characteristics of the boundary curve s and the correlation of the properties to the region's semantic identity. Then we discuss the imaging process of a simplified 3D scene model and sh ow how it explains the empirical observations. We further show that the ob served properties suffice to characterize the gist of scenery images, prop ose a generative parametric graphical model, and use it to learn and gener ate semantic sketches of new images, which indeed look like those associat ed with natural scenery.

Joint work with Michael Lindenbau m. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012260 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110111T143000 DTEND;TZID="Asia/Jerusalem":20110111T163000 DTSTAMP;TZID="Asia/Jerusalem":20110111T143000 FREEBUSY;FBTYPE=BUSY:20110111T143000/20110111T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Nayantara Bhatnagar about Recons truction in Trees at 2011-01-11 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011660 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110112T122000 DTEND;TZID="Asia/Jerusalem":20110112T142000 DTSTAMP;TZID="Asia/Jerusalem":20110112T122000 FREEBUSY;FBTYPE=BUSY:20110112T122000/20110112T142000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Oren Weimann (Weizmann I nstitute) about Theory Seminar: Fault-tolerant Shortest Paths and Minimum Spanning Trees at 2011-01-12 12:20:00 DESCRIPTION;LANGUAGE=en-US:A fundamental problem in dynamic graphs is the recovery of structural information in a network whose edges occasionally f ail. In a failure event, some subset of edges D are deleted and we want to quickly understand the structure of the surviving network G \ D. In par ticular, I will discuss the problem of quickly recovering shortest paths a nd a minimum spanning tree in G \ D.

Let G = (V,E) be a direct ed edge-weighted graph and let P be a shortest path from s to t in G. The "replacement paths" problem asks to compute, for every edge e on P, the sh ortest s-to-t path that avoids e. Apart from approximation algorithms and algorithms for special graph classes, the naive solution to this problem - - removing each edge e on P one at a time and computing the shortest s-to- t path each time -- was surprisingly the only known solution for directed weighted graphs, even when the weights are integrals. In particular, altho ugh the related shortest paths problem has benefited from fast matrix mult iplication, the replacement paths problem has not, and still required cubi c time.

For an n-vertex graph with integral edge-lengths in {- M,...,M}, I will describe an O(Mn^{2.584}) time algorithm that uses fast m atrix multiplication and is sub-cubic for appropriate values of M. We also show how to construct a "distance sensitivity oracle" in the same time bo unds. A query (u,v,e) to this oracle requires sub-quadratic time and retur ns the length of the shortest u-to-v path that avoids the edge e. Our resu lts also apply for avoiding multiple edges and for avoiding vertices rathe r than edges.

The talk is based on a FOCS'10 paper with Raphael Yuster. If time permits, I will also discuss a new result with Shiri Chec hick and David Peleg for maintaining a minimum spanning tree of a graph su bject to edge failures. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012320 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110113T143000 DTEND;TZID="Asia/Jerusalem":20110113T163000 DTSTAMP;TZID="Asia/Jerusalem":20110113T143000 FREEBUSY;FBTYPE=BUSY:20110113T143000/20110113T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Joel Lanir SPECIAL LECTURE about Designing presentations tools for multiple and high-resolution displays a t 2011-01-13 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400011680 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110116T113000 DTEND;TZID="Asia/Jerusalem":20110116T133000 DTSTAMP;TZID="Asia/Jerusalem":20110116T113000 FREEBUSY;FBTYPE=BUSY:20110116T113000/20110116T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Prof. Arie Kaufman (CS, SUNY ) about Pixel Club: Immersive Visualization of Large Datasets at 2011-01-1 6 11:30:00 DESCRIPTION;LANGUAGE=en-US:Scientists, engineers and physicians are now co nfronted with a fire hose of data. Immersive visualization environments pr ovide these users with a novel way of interacting and reasoning with large datasets. They allow them to utilize the entirety of their visual bandwid th, effectively engulfing the user in the data and enabling collaborative interaction. We present a custom-built 5-wall Cave environment, called the Immersive Cabin (IC). It is driven by a GPU cluster for both computation and 3D stereo rendering. We also propose a conformal deformation rendering pipeline for the visualization of datasets on partially-immersive platfor ms. Combined with a range of interaction and navigation tools, our system can support numerous interactive applications of large datasets. Several d emonstrations include architectural visualization, urban planning, medical visualization, simulation and rendering of physical phenomena, and entert ainment.

Current visualization displays, however, have not k ept up with the explosive growth in data size and high resolution, which i s beginning to match the resolution of the visuals that surround us in dai ly life. To ameliorate this challenge, we have developed a life-like, real istic immersion into the petascale data, appropriately called The RealityD eck. It is a one-of-a kind pioneering G-pixel immersive and collaborative display system – a unique assembly of high-resolution display panels, GP U cluster, sensors, networking, computer vision, and human-computer intera ction technologies. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400012370 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110117T183000 DTEND;TZID="Asia/Jerusalem":20110117T203000 DTSTAMP;TZID="Asia/Jerusalem":20110117T183000 FREEBUSY;FBTYPE=BUSY:20110117T183000/20110117T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Boaz Goldstein (Mic rosoft) about Haifux Club: A FOSS Yankee in Microsoft's Court at 2011-01- 17 18:30:00 DESCRIPTION;LANGUAGE=en-US:A year ago I started working for a small multi- national software giant named Microsoft. What I found is a company with ra ther surprising and odd corporate culture and habbits. This lecture will t ry to convey What microsoft is like on the inside. Disclaimer: anything said in this lecture is my opinion alone and is not affiliated with micro soft in any way. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400012300 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110119T122000 DTEND;TZID="Asia/Jerusalem":20110119T142000 DTSTAMP;TZID="Asia/Jerusalem":20110119T122000 FREEBUSY;FBTYPE=BUSY:20110119T122000/20110119T142000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Dana Moshkovitz (MIT) ab out Theory Seminar: Hardness of Approximately Solving Linear Equations Ove r Reals at 2011-01-19 12:20:00 DESCRIPTION;LANGUAGE=en-US:We consider the problem of approximately solvin g a system of homogeneous linear equations over reals, where each equati on contains at most three variables. Since the all-zero assignment alw ays satisfies all the equations exactly, we restrict the assignments to be ``non-trivial". Here is an informal statement of our result: it is NP -hard to distinguish whether there is a non-trivial assignment that sati sfies 1-\delta fraction of the equations or every non-trivial assignment fails to satisfy a constant fraction of the equations with a ``margin" of \Omega(\sqrt{\delta}).

Unlike the well-studied case of lin ear equations over finite fields, for equations over reals, the best app roximation algorithm known (SDP-based) is the same no matter whether the number ofvariables per equation is two or three.

Our re sult is motivated by the following potential approach to proving The Uni que Games Conjecture: 1. Prove the NP-hardness of solving approximate li near equations over reals, for the case of three variables per equation (we prove this result). 2. Prove the NP-hardness of the problem for the case of two variablesper equation, possibly via a reduction from the thr ee variable case. 3. Prove the Unique Games Conjecture.

An interesting feature of our result is that it shows NP-hardness result that matches the performance of a non-trivial SDP-algorithm. Indeed, th e Unique Games Conjecture predicts that an SDP-based algorithm is optima l for a huge class of problems (e.g. all CSPs by Raghavendra's result).

This is joint work with Subhash Khot. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 401 UID:111ec4410400012390 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110119T150000 DTEND;TZID="Asia/Jerusalem":20110119T170000 DTSTAMP;TZID="Asia/Jerusalem":20110119T150000 FREEBUSY;FBTYPE=BUSY:20110119T150000/20110119T170000 SUMMARY;LANGUAGE=en-US:The msc talk by Boaz Brickner about Methods For Rec ognition By Graphical Style And Style Synthesis Using Local Analysis at 20 11-01-19 15:00:00 DESCRIPTION;LANGUAGE=en-US:Standard image classification algorithms classi fy an image by its content. Sometimes, we don't care about the image conte nt and want to classify images by style. For example, if we have a set of paintings and we want to divide them to groups by their painting style or painter, we don't care if a car or a horse was painted. Another usage for classifying by style is when we have a painting and we are not sure whethe r it was really painted by a specific painter, and we want to check whethe r its style matches the painter's style. We call this image classification problem recognition by graphical style. A related problem is applying an image style to an image. An algorithm that can do that can repaint an ima ge using a specific painter style as if the painter painted this image. Th e idea is to extract the style of an image (or images) and recreate a diff erent image so it would still have the original image's content but with t he extracted image style. We call this problem graphical style synthesis. Both these problems require local analysis of the images instead of globa l analysis that is done when we look for the image content instead of the image style. We study these problems and develop machine learning based a lgorithm for recognition by graphical style and heuristic algorithms for g raphical style synthesis. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012360 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110123T130000 DTEND;TZID="Asia/Jerusalem":20110123T150000 DTSTAMP;TZID="Asia/Jerusalem":20110123T130000 FREEBUSY;FBTYPE=BUSY:20110123T130000/20110123T150000 SUMMARY;LANGUAGE=en-US:The phd talk by Amir Vaxman about General Technique s for Interpolation, Reconstruction, and Morphing of Polyhedral Surfaces a t 2011-01-23 13:00:00 DESCRIPTION;LANGUAGE=en-US:The topic of shape reconstruction is a major re search branch of geometric processing. The two major problems which are commonly researched are reconstruction of two and three-dimensional object s from point sets, and reconstruction from cross-sections. This research focuses on the latter, which has received a substantial amount of atten tion in the last three decades, and which has specifically gained consid erable momentum in the last few years. We propose two solutions for the p roblem of reconstruction from cross-sections: A geometric approach, util izing the three-dimensional straight skeleton, and a variational shape app roach, which seeks to discover the best interpolating surface in terms o f an energy functional. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400011930 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110125T113000 DTEND;TZID="Asia/Jerusalem":20110125T133000 DTSTAMP;TZID="Asia/Jerusalem":20110125T113000 FREEBUSY;FBTYPE=BUSY:20110125T113000/20110125T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Shai Furman (EE, Technion) a bout Pixel Club: Multidimensional Image Representation and Processing Moti vated by Human Vision at 2011-01-25 11:30:00 DESCRIPTION;LANGUAGE=en-US:A biological model of visual information repres entation is adopted. Images are represented accordingly in a multidimensio nal space that incorporates the well investigated dimensions of intensity, color and spatio-temporal frequency. The model is extended to incorporate additional, less investigated dimensions such as curvature, size and dept h (for example - from binocular disparity). Along these and other dimensio ns, that are yet to be discovered, the human visual system (HVS) enhances and emphasizes important image attributes by adaptation and nonlinear filt ering.

The non-linear Automatic Gain Control (AGC) model of processing along the visual dimensions is presented together with its biol ogical foundation. A biologically-motivated artificial neural network (ANN ) implementation is presented as an example. The model is analyzed for its SNR characteristics. Several inputs and responses are considered and impl emented along the visual dimensions of curvature, size and depth. The resu lts are compared with those of psychophysical experiments, exhibiting good reproduction of visual illusions. Finally, examples of applications of th e AGC model in image processing and computer vision are presented. These i nclude HDR images, enhanced edge detection and curve completion due to occ lusion.

Implementing the generic neural AGC model along all vis ual dimensions constitutes a universal, parsimonious and unified model tha t proposes how our visual system processes visual information along its va rious dimensions, before the later stage of sequential “visual routines is implemented. This approach may lead to the development of a metric f or calculation of distance between images, and facilitate the execution of important tasks, such as recognition and classification. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400012380 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110126T113000 DTEND;TZID="Asia/Jerusalem":20110126T133000 DTSTAMP;TZID="Asia/Jerusalem":20110126T113000 FREEBUSY;FBTYPE=BUSY:20110126T113000/20110126T133000 SUMMARY;LANGUAGE=en-US:The phd talk by Royi Ronen about Models and Methods for Social Networks Automation at 2011-01-26 11:30:00 DESCRIPTION;LANGUAGE=en-US:This research is concerned with novel Web-relat ed data management scenarios. In particular, we consider database proble ms which arise from social networks. In the talk, we will overview the m ain results of this research. In the first part, we introduce the Que ry Network model, a basic model for social network automation using quer ies, with its evaluation algorithms and related experiments. We also dis cuss extensions for the model, and present a few theoretical results. In the second part, we discuss social networks automation using Protocols for Social Networks, which coordinate consistency-preserving decisions in the network. Methods for the concurrent execution of multiple protoco l instances, which relax the traditional concurrency control isolation requirement, are presented. In the third part, we describe XPath^L lan guage and system, which combines XPath, an XML query language, and Data log, by introducing an XPath predicate to the Datalog logic-based formal ism. We overview algorithms and related theoretical results. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012430 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110126T122000 DTEND;TZID="Asia/Jerusalem":20110126T142000 DTSTAMP;TZID="Asia/Jerusalem":20110126T122000 FREEBUSY;FBTYPE=BUSY:20110126T122000/20110126T142000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Oded Regev (Tel-Aviv Uni versity) about Theory Seminar: Quantum One-Way Communication can be Expon entially Stronger Than Classical Communication at 2011-01-26 12:20:00 DESCRIPTION;LANGUAGE=en-US:In STOC 1999, Raz presented a (partial) functio n for which there is a quantum protocol communicating only $O(\log n)$ qub its, but for which any classical (randomized, bounded-error) protocol requ ires $\poly(n)$ bits of communication. That quantum protocol requires two rounds of communication. Ever since Raz's paper it was open whether the sa me exponential separation can be achieved with a quantum protocol that use s only one round of communication. In other words, can quantum one-way com munication be exponentially stronger than classical two-way communication? Here we settle this question in the affirmative.

Based on jo int work with Bo'az Klartag.

NOTE: This talk is about lower b ounds for *classical* communication complexity, so it does not require any knowledge in quantum communication complexity. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012460 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110126T140000 DTEND;TZID="Asia/Jerusalem":20110126T160000 DTSTAMP;TZID="Asia/Jerusalem":20110126T140000 FREEBUSY;FBTYPE=BUSY:20110126T140000/20110126T160000 SUMMARY;LANGUAGE=en-US:The msc talk by Neer Roggel about Anonymous Routing in Mobile Ad Hoc Networks at 2011-01-26 14:00:00 DESCRIPTION;LANGUAGE=en-US:A wireless, mobile, ad hoc network (MANET) is a network in which mobile nodes do not rely on the existence of fixed infr astructure mediation devices, but rather communicate with one another dir ectly. Under certain scenarios, parties in a MANET may wish to remain uni dentified, in order to forestall retaliation by an attacker. In this talk , we discuss mechanisms for anonymous routing in MANETs. First, we note t hreats to anonymity in MANETs, derive a general adversary model and use i t to define anonymity in a formal yet intuitive manner. Then, we use our framework to analyze the anonymity of existing routing algorithms under v arious identity agreements which we define. Lastly, we briefly compare our results with those of others. The talk is self-contained. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012510 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110127T143000 DTEND;TZID="Asia/Jerusalem":20110127T163000 DTSTAMP;TZID="Asia/Jerusalem":20110127T143000 FREEBUSY;FBTYPE=BUSY:20110127T143000/20110127T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Tal Sobol Shikler about Analysis of affective behaviour: Non-verbal Speech - Feature Extraction, Multi-cla ss & Multi-label Classification, and Generalisation of the Technology a t 2011-01-27 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012440 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110130T130000 DTEND;TZID="Asia/Jerusalem":20110130T150000 DTSTAMP;TZID="Asia/Jerusalem":20110130T130000 FREEBUSY;FBTYPE=BUSY:20110130T130000/20110130T150000 SUMMARY;LANGUAGE=en-US:The phd talk by Gadi Aleksandrowicz about Enumerati on of Lattice Animals at 2011-01-30 13:00:00 DESCRIPTION;LANGUAGE=en-US:A Lattice Animal is a set of edge-connected cel ls in a given lattice. For example, the Tetris game is played with Lattic e Animals with 4 cells in the two-dimensional orthogonal lattice. The e numeration of Lattice Animals is a long-time standing problem, arising in all of recreational mathematics, discrete geometry, and statistical phys ics. In this talk I will discuss a generalization of Redelmeier's algori thm to the enumeration of polyominoes that lie on any structural (repetit ve) lattice. I will also present a bijection between polyominoes and pe rmutations, and describe the enumeration of several families of polyomino es through couting classes of permutations that avoid some sets of "forbi dden patterns". In addition, I will apply a transfer-matrix method to po lyominoes on a twisted cylinder in order to derive the generating functio ns of the sequence that enumerates these polyominoes. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400012530 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110131T183000 DTEND;TZID="Asia/Jerusalem":20110131T203000 DTSTAMP;TZID="Asia/Jerusalem":20110131T183000 FREEBUSY;FBTYPE=BUSY:20110131T183000/20110131T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Eli Billauer about Haifux Club: Root on NFS: Running Linux on a Diskless Computer at 2011- 01-31 18:30:00 DESCRIPTION;LANGUAGE=en-US:A motherboard, a CPU, and a memory stick. Add a fan and a power supply, and you have a little computer which boots from n etwork and keeps all its data on a hosting computer's disk through NFS. Th e shopping list for new components goes as low as 600 NIS, which makes thi s an attractive solution in cases where a dedicated computer is useful, be it for mission-critical applications, cases where the hardware may be dam aged or stolen, or because the OS is bound to crash often (kernel hacking) .

This lecture is a walktrough of my own recent expe rience of setting up a diskless computer, using cobbler for coordinatin g the PXE boot. It will involve the gory technical details as well as m y own insights after making a fullblown Linux system run without any local storage. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400012480 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110201T143000 DTEND;TZID="Asia/Jerusalem":20110201T163000 DTSTAMP;TZID="Asia/Jerusalem":20110201T143000 FREEBUSY;FBTYPE=BUSY:20110201T143000/20110201T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Yoav Etsion about Task Superscal ar Multiprocessors at 2011-02-01 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012330 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110202T123000 DTEND;TZID="Asia/Jerusalem":20110202T143000 DTSTAMP;TZID="Asia/Jerusalem":20110202T123000 FREEBUSY;FBTYPE=BUSY:20110202T123000/20110202T143000 SUMMARY;LANGUAGE=en-US:The msc talk by Muhamad Mjadleh about Accelerating CIFS Over Satellite Networks at 2011-02-02 12:30:00 DESCRIPTION;LANGUAGE=en-US:CIFS is the underlying protocol in Windows OS's for disk shares access such as copying files from and to remote sta tions in the LAN, exploring disk shares and sending jobs to a LAN print er. CIFS is also implemented for Linux/Unix OSs in SAMBA. The CIFS prot ocol is chatty and suffers from redundancy in messages even in simple a nd basic disk shares accesses. CIFS's ineffectiveness is intensified in satellite networks in which the round trip delay is large. This wo rk presents a proxy-based approach to CIFS acceleration in satellite ne tworks. The acceleration is done at the level of the application layer without any modification neither to the TCP layer nor to the server and the client. We studied and observed CIFS protocol behavior and develop ed a solution that is fitted to the protocol. The solution is ligh t-weight, it requires little memory space, is transparent to both clien t machines and servers, and it fits into existing VSAT technologies aim ed at developing countries. We adapted several techniques and developed algorithms to lessen the number of round trip interactions required to perfrom a single user operation. Performance evaluation shows lat ency improvements by a factor of two in many practical settings. In add ition, bandwidth consumption is reduced. The talk is self-contained. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012540 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110202T160000 DTEND;TZID="Asia/Jerusalem":20110202T180000 DTSTAMP;TZID="Asia/Jerusalem":20110202T160000 FREEBUSY;FBTYPE=BUSY:20110202T160000/20110202T180000 SUMMARY;LANGUAGE=en-US:The msc talk by Anna Kaplun about Density-Driven Pu blish Subscribe Service for Mobile Ad-Hoc Networks at 2011-02-02 16:00:00 DESCRIPTION;LANGUAGE=en-US:We study a publish/subscribe service for mobile ad-hoc networks. Mobile Ad-Hoc Networks (MANETs) are formed by a collect ion of mobile nodes, each equipped with wireless communication capabilitie s, without relying on any fixed infrastructure or centralized administrati on. A publish/subscribe system is responsible for delivering data from a source to interested users. A user expresses interest in receiving certain data by submitting a predicate about corresponding data content. In th is talk, we introduce the density driven virtual topography that is used t o route the messages through the network. Then, we overview two variants o f a publish/subscribe service that is based on the virtual topography. Las tly, we compare our results with the tree-based publish/subscribe mechanis m and an optimized flooding based scheme. Our results show that the densit y driven approach is superior to the others in most scenarios for most met rics. The talk is self-contained. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012520 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110208T143000 DTEND;TZID="Asia/Jerusalem":20110208T163000 DTSTAMP;TZID="Asia/Jerusalem":20110208T143000 FREEBUSY;FBTYPE=BUSY:20110208T143000/20110208T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Michal Aharon about The PARIS Al gorithm for Determining Latent Topics at 2011-02-08 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012450 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110214T183000 DTEND;TZID="Asia/Jerusalem":20110214T203000 DTSTAMP;TZID="Asia/Jerusalem":20110214T183000 FREEBUSY;FBTYPE=BUSY:20110214T183000/20110214T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Amichay Peretz Klop shtock about Haifux Club: Encryption - Alice, Bob and Co. - Amichay Peretz Klopshtock at 2011-02-14 18:30:00 DESCRIPTION;LANGUAGE=en-US:In this lecture I'll talk about the encryption methods that were common through history, How to use them, what their weak nesses are and how you can break them. During the lecture I will give exam ples. The audience is requested to bring writing equipment and paper. N o previous background required. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400012490 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110216T133000 DTEND;TZID="Asia/Jerusalem":20110216T153000 DTSTAMP;TZID="Asia/Jerusalem":20110216T133000 FREEBUSY;FBTYPE=BUSY:20110216T133000/20110216T153000 SUMMARY;LANGUAGE=en-US:The msc talk by Gilad Kutiel about Cost-Aware Live Migration of Services in the Cloud at 2011-02-16 13:30:00 DESCRIPTION;LANGUAGE=en-US:Live migration of virtual machines is an import ant component of the emerging cloud computing paradigm. While live migra tion provides extreme versatility of management, it comes at a price of degraded service performance during migration. The bulk of studies devot ed to live migration of virtual machines focus on the duration of the copy phase as a primary metric of migration performance. While shorter down times are clearly desirable, the pre-copy phase imposes an overhead on t he infrastructure that may result in severe performance degradation of the migrated and collocated services offsetting the benefits accrued through live migration. We observe that there is a non-trivial trade-off between minimizing the copy phase duration and maintaining an acceptable quality of service during the pre-copy phase, and introduce a new model to quant ify this trade-off. We then show that using our model, an optimal migr ation schedule can be efficiently calculated. Finally, we simulate, usin g real traces, live migrations of a virtual machine running a web server and compare the migration cost using our algorithm and commonly used live -migration methods. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012580 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110216T153000 DTEND;TZID="Asia/Jerusalem":20110216T173000 DTSTAMP;TZID="Asia/Jerusalem":20110216T153000 FREEBUSY;FBTYPE=BUSY:20110216T153000/20110216T173000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Yishai Shimoni (An drea Califano's Lab, Center for Computational Biology and Bioinformatics ( C2B2), Columbia University, New York) about Bioinformatics Forum: Stochast ic Effects in Viral-infected Dendritic Cells Lead to Efficient Immune Res ponse Activation at 2011-02-16 15:30:00 DESCRIPTION;LANGUAGE=en-US:When monocyte-derived human dendritic cells (DC s) are infected by Newcastle disease virus, the virus is known to be det ected by RIG-I proteins, which induces interferon production. Interferon activates a host of genes, including the gene coding RIG-I. Single cell measurements is DCs show large cell to cell variation of 3-4 orders of magnitude at 6-10 hours after infection. In order analyze early times af ter infection, when reliable direct single cell data cannot be obtained, an agent-based mathematical model was developed. The model was correlat ed with biochemical time-course measurements of the levels of IFNB1 and DDX58 (RIG-I). Simulations showed that a high level of variation and the presence at early times of a small number of early responder cells is nec essary for explaining the experimental data as well as for efficient and controlled activation of the IFNB1-DDX58 positive feedback loop. The mo del generated testable predictions that were confirmed by single cell ex periments. The results suggest that large cell-to-cell response variatio n plays a significant role in the early innate immune response, and that the variability is in fact essential to the efficient activation of the IFNB1 based feedback loop. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400012550 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110223T110000 DTEND;TZID="Asia/Jerusalem":20110223T130000 DTSTAMP;TZID="Asia/Jerusalem":20110223T110000 FREEBUSY;FBTYPE=BUSY:20110223T110000/20110223T130000 SUMMARY;LANGUAGE=en-US:The cggc talk by Alla Sheffer (Computer Science, Un iv. of British Columbia, Vancouver, Canada) about CGGC Seminar: Space-T ime Reconstruction - Understanding Motion at 2011-02-23 11:00:00 DESCRIPTION;LANGUAGE=en-US:As research on space-time reconstruction mature s, we should ask ourselves what information we use to correctly leverage the temporal or motion component of the data.

In my talk I wi ll discuss several possible motion priors and their impact on the recons truction.

Specifically, I will address the observation that m ost changes in shape are gradual, in both intrinsic and Euclidean sense. I will discuss the impact of this observation on the interpretation of shape over time and present a reconstruction method that uses this obse rvation to correctly reconstruct shape and motion in the presence of sel f-contacts. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 5 UID:111ec4410400012620 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110223T130000 DTEND;TZID="Asia/Jerusalem":20110223T150000 DTSTAMP;TZID="Asia/Jerusalem":20110223T130000 FREEBUSY;FBTYPE=BUSY:20110223T130000/20110223T150000 SUMMARY;LANGUAGE=en-US:The msc talk by Uri Avraham about ABC - A New Frame work for Block Ciphers at 2011-02-23 13:00:00 DESCRIPTION;LANGUAGE=en-US:We suggest a new framework for block ciphers na med Advanced Block Cipher, or shortly ABC. ABC has additional non-secret parameters that ensure that each call to the underlying block cipher uses a different pseudo-random permutation. It therefore ensures that attacks that require more than one block encrypted under the same secret permutati on cannot apply. In particular, this framework protects against dictionary attacks, and differential and linear attacks, and eliminates weaknesses o f ECB and CBC modes. This new framework shares a common structure with HAI FA, and can share the same logic with HAIFA compression functions. We anal yze the security of several modes of operation for ABCs block ciphers, and suggest a few instances of ABCs. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012630 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110227T140000 DTEND;TZID="Asia/Jerusalem":20110227T160000 DTSTAMP;TZID="Asia/Jerusalem":20110227T140000 FREEBUSY;FBTYPE=BUSY:20110227T140000/20110227T160000 SUMMARY;LANGUAGE=en-US:The colloq talk by Prof. David Abramson SPECIAL LEC TURE note unusual hour about e-Science: Are we there yet? at 2011-02-27 14 :00:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012610 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110228T183000 DTEND;TZID="Asia/Jerusalem":20110228T203000 DTSTAMP;TZID="Asia/Jerusalem":20110228T183000 FREEBUSY;FBTYPE=BUSY:20110228T183000/20110228T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Vadim Eisenberg (IB M HRL and CS Technion) about Haifux Club: UniversAAL: Open Source platform for Ambient Assisted Living and Smart Home Environment at 2011-02-28 1 8:30:00 DESCRIPTION;LANGUAGE=en-US:I will present two presentations about an EU FP 7 IP project I work on, UniversAAL - http://universaal.org/. The goal o f the project is to develop an open source platform for Ambient Assisted L iving (AAL). Ambient Assisted Living is a kind of Smart Home environmen t for elderly people - for example, a house equipped with different sen sors and in which different devices, sensors and home appliances are ne tworked together and managed by software applications. In addition, the platform could be used for monitoring of chronic diseases, providing heal thcare services at home, supporting people with disabilities. I will pr esent the project and will talk about technologies involved in the proj ect: OSGi, Middleware, (Semantic) SOA, Security, Android, Living Labs.

About me: I work in IBM, IT for Healthcare & Life Sciences grou p, and do my M.Sc. studies at the Technion, Computer Science Department. M y research interests are Semantic Web, Software Engineering and Programmin g Languages. I made my first Open Source contribution two months ago - htt p://d2rqupdate.cs.technion.ac.il/, Apache 2.0 license (on my personal time , it is not related to IBM). I developed D2RQ/Update and D2R Server/Update prototype extensions to a popular RDF-to-RDB mapping platform - D2RQ http ://www4.wiwiss.fu-berlin.de/bizer/d2rq/ (it is related to Semantic Web tec hnologies). ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400012500 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110301T143000 DTEND;TZID="Asia/Jerusalem":20110301T163000 DTSTAMP;TZID="Asia/Jerusalem":20110301T143000 FREEBUSY;FBTYPE=BUSY:20110301T143000/20110301T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Leonid Barenboim about Determini stic Distributed Vertex Coloring in Polylogarithmic Time at 2011-03-01 14: 30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012560 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110302T113000 DTEND;TZID="Asia/Jerusalem":20110302T133000 DTSTAMP;TZID="Asia/Jerusalem":20110302T113000 FREEBUSY;FBTYPE=BUSY:20110302T113000/20110302T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Amit Golander (IBM Research) abo ut ceClub: IBM's PowerEN Developer Cloud: Fertile Ground for Academic Rese arch at 2011-03-02 11:30:00 DESCRIPTION;LANGUAGE=en-US:IBM's newest technology, the Power Edge of Netw ork (PowerEN) processor, merges network and server attributes to create a new class of wire-speed processor. PowerEN is a hybrid computer that emplo ys: massive multithreading capabilities, integrated I/O and unique special -purpose accelerators for compression, cryptography, pattern matching, XML and Network processing.

As a novel architecture, the PowerE N processor offers fertile ground for research. It can facilitate the deve lopment of faster applications in many fields of computer science such as Networking, Cryptography, Virtualization, Bioinformatics, SOA, and Systems . IBM encourages academic research and has set up the "PowerEN Developer C loud" infrastructure to allow it. Israeli universities are of the first to perform research on this global infrastructure. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400012650 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110302T123000 DTEND;TZID="Asia/Jerusalem":20110302T143000 DTSTAMP;TZID="Asia/Jerusalem":20110302T123000 FREEBUSY;FBTYPE=BUSY:20110302T123000/20110302T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Yahav Nussbaum (Tel-Aviv University) about Theory Seminar: Improved Minimum Cuts and Maximum Flows in Undirected Planar Graphs at 2011-03-02 12:30:00 DESCRIPTION;LANGUAGE=en-US:The problem of finding the minimum s-t cut betw een a source s and a sink t in a graph is a well-studied problem in comput er science. Finding the minimum s-t cut in a planar graph has applications in many fields - from traffic design to computer vision.

Th e geometrical duality between a minimum s-t cut in an undirected planar gr aph and cycle on the plane that separates t from s was first used by Itai and Shiloach to solve the minimum cut problem, and more efficient algorith ms for the problem using the same approach followed.

We use this approach to show an algorithm that finds the minimum s-t cut in an n- vertices undirected planar graph in O(n log log n) time. We then use the i mproved minimum cut algorithm to solve related problems, including the max imum flow problem in undirected planar graphs, within the same time bound.

Using a similar technique we obtain a dynamic data structure that allows edge capacity changes and answers minimum cut queries between any pair of vertices in O(n^(2/3) polylog(n)) time for undirected planar graphs.

Joint work with Giuseppe F. Italiano, Piotr Sankowsk i, and Christian Wulff-Nilsen ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400012690 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110302T133000 DTEND;TZID="Asia/Jerusalem":20110302T153000 DTSTAMP;TZID="Asia/Jerusalem":20110302T133000 FREEBUSY;FBTYPE=BUSY:20110302T133000/20110302T153000 SUMMARY;LANGUAGE=en-US:The msc talk by Alexandra Faynburd about Short-term turning points forecasting in financial time series at 2011-03-02 13:30:0 0 DESCRIPTION;LANGUAGE=en-US:We consider the problem of forecasting turning points in time series and develop an autoregressive prediction algorithm, that relies on a novel turning point indicator and support vector regres sion. The algorithm is analyzed and compared to existing methods in the c ontext of financial forecasting. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012670 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110306T150000 DTEND;TZID="Asia/Jerusalem":20110306T170000 DTSTAMP;TZID="Asia/Jerusalem":20110306T150000 FREEBUSY;FBTYPE=BUSY:20110306T150000/20110306T170000 SUMMARY;LANGUAGE=en-US:The msc talk by Nadav Amit about vIOMMU: Efficient IOMMU Emulation at 2011-03-06 15:00:00 DESCRIPTION;LANGUAGE=en-US:Direct device assignment, where a guest virtual machine directly interacts with an I/O device without host intervention, is appealing, because it allows an unmodified (non-hypervisor-aware) gue st to achieve near-native performance. But device assignment for unmodifi ed guests suffers from two serious deficiencies: (1) it requires pinning of all the guest's pages, thereby disallowing memory overcommitment, and (2) it exposes the guest's memory to buggy device drivers. We solve th ese problems by designing, implementing, and exposing an emulated IOMMU ( vIOMMU) to the unmodified guest. We employ two novel optimizations to mak e vIOMMU perform well: (1) waiting a few milliseconds before tearing down an IOMMU mapping in the hope it will be immediately reused (``optimistic teardown''), and (2) running the vIOMMU on a sidecore, and thereby enabl ing for the first time the use of a sidecore by unmodified guests. Both o ptimizations are highly effective in isolation. The former allows bare-me tal to achieve 100% of a 10Gbps line rate. The combination of the two all ows an unmodified guest to do the same. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012730 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110308T113000 DTEND;TZID="Asia/Jerusalem":20110308T133000 DTSTAMP;TZID="Asia/Jerusalem":20110308T113000 FREEBUSY;FBTYPE=BUSY:20110308T113000/20110308T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Guy Gilboa (Mathematics, UCL A) about Pixel Club: Nonlocal Variational Methods for Image Processing at 2011-03-08 11:30:00 DESCRIPTION;LANGUAGE=en-US:Variational and PDE-based methods have been ext ensively used for vision and image-processing tasks such as denoising, seg mentation, inpainting, optical flow and more. An underlying assumption is the (piecewise) local correlation between pixels in typical images. Imag es also exhibit nonlocal correlations in repetitive structures and texture s. A coherent mathematical framework for nonlocal regularization will be p resented in this talk, which stems from graph theory. The notions of nonlo cal derivatives, diffusion processes and regularizers will be defined and used for various image processing applications. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012740 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110308T143000 DTEND;TZID="Asia/Jerusalem":20110308T163000 DTSTAMP;TZID="Asia/Jerusalem":20110308T143000 FREEBUSY;FBTYPE=BUSY:20110308T143000/20110308T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Hilary W. Putnam about The Curio us Story of Quantum Logic at 2011-03-08 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012570 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110309T113000 DTEND;TZID="Asia/Jerusalem":20110309T133000 DTSTAMP;TZID="Asia/Jerusalem":20110309T113000 FREEBUSY;FBTYPE=BUSY:20110309T113000/20110309T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Erven Rohou (INRIA Rennes, ALF R esearch Group) about ceClub: Portability and Performance of Applications i n the Manycore Era at 2011-03-09 11:30:00 DESCRIPTION;LANGUAGE=en-US:For many years, the frequency of microprocessor s has grown at an exponential pace (from 20 MHz in 1990 to 2 GHz in 2000). Since 2002 however, despite considerable effort, the frequency has been p lateauing. Advances in technology and micro-architecture now translate int o more parallelism. The consequences on the software industry are dramatic : most existing software has been designed with a sequential model in mind , and even parallel applications contain residual sequential sections. The performance of these applications will ultimately be driven by the perfor mance of the sequential part, as stated by Amdahl's law.

For this reason we envision that future architectures will consist of many si mple cores, but also a few very complex cores. The performance of applicat ions is unlikely to automatically increase with new generations of future processors, as it used to happen in the past. Even plain portability is at risk for applications developed with a specific model of parallelism in m ind. This talk proposes directions to address the problem of portability a nd performance of applications on future manycore processors, taking into account both parallel and residual sequential code sections. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400012660 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110309T123000 DTEND;TZID="Asia/Jerusalem":20110309T143000 DTSTAMP;TZID="Asia/Jerusalem":20110309T123000 FREEBUSY;FBTYPE=BUSY:20110309T123000/20110309T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Pavel Hrubes (Institute of Advanced Studies, Princeton) about Theory Seminar: How Much Commutativi ty is Needed to Prove Polynomial Identities? at 2011-03-09 12:30: 00 DESCRIPTION;LANGUAGE=en-US:The so called Extended Frege system is one of t he most natural propositional proof systems. Whereas we believe that there exist tautologies which require exponential Extended Frege proofs, the be st known lower bound is linear. To find even a modestly superlinear lower bound is a challenging open problem. I will discuss a possible approach to this question, which is based on counting the number of commutativity axi oms in an Extended Frege proof, which in turn can be phrased as an algebra ic problem about non-commutative polynomials. We will see that this line o f reasoning generates new difficult questions, rather than answering the o riginal one. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400012700 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110314T143000 DTEND;TZID="Asia/Jerusalem":20110314T163000 DTSTAMP;TZID="Asia/Jerusalem":20110314T143000 FREEBUSY;FBTYPE=BUSY:20110314T143000/20110314T163000 SUMMARY;LANGUAGE=en-US:The CSpecial Talk talk by Prof. Roman Vitenberg (De partment of Informatics, University of Oslo) about CSpecial Talk: Scala ble Overlay Design for Topic-Based Publish/Subscribe Systems and Fast Ov erlay Construction Algorithms at 2011-03-14 14:30:00 DESCRIPTION;LANGUAGE=en-US:Pub-sub is a paradigm for asynchronous communic ation that is commonly used in a great variety of industrial application s, such as news tickers, delivery of financial data, Military applicatio ns, and many others. While client-server communication still remains th e prevailing implementation paradigm for pub-sub, its limitations for large-scale applications are widely recognized. Major industrial players such as Google and Tibco are recognizing the potential of cooperative o verlays and starting to replace legacy centralized architectures. Howeve r, the lack of adequate technologies hinders this industrial shift. In this talk, I will consider the problem of designing a scalable o verlay network to support decentralized topic-based pub-sub communicatio n. I will present the overview of the state-of-the-art in this area and our recent results, both analytical and experimental, based on the ICDCS '10 and ICDCS'11 publications. These results (a) improve the runtime com plexity of existing algorithms, (b) address the challenge of incremental construction by introducing the overlay join problem, and (c) allow for parallelized overlay construction while retaining the high quality of t he centralized design. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012770 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110314T183000 DTEND;TZID="Asia/Jerusalem":20110314T203000 DTSTAMP;TZID="Asia/Jerusalem":20110314T183000 FREEBUSY;FBTYPE=BUSY:20110314T183000/20110314T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Guy Keren about Hai fux Club: The Story of Alice and Bob - the I/O Requests (Part I) at 2011-0 3-14 18:30:00 DESCRIPTION;LANGUAGE=en-US:In this story, we'll follow the life story of a lice - a file-systemized I/O request, and bob - a raw-device I/O request, from their birth, until they reach heaven (the disk or the USB camera or.. .). In addition, we will cover some system parameters that affect I/O requ ests, the buffer cache, disk I/O schedulers and tools used to track and co unt I/O (including - why is process-based I/O accounting so tricky). ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400012750 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110315T143000 DTEND;TZID="Asia/Jerusalem":20110315T163000 DTSTAMP;TZID="Asia/Jerusalem":20110315T143000 FREEBUSY;FBTYPE=BUSY:20110315T143000/20110315T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Shmuel Onn about The Graver Comp lexity of Integer Programming at 2011-03-15 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012680 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110315T160000 DTEND;TZID="Asia/Jerusalem":20110315T180000 DTSTAMP;TZID="Asia/Jerusalem":20110315T160000 FREEBUSY;FBTYPE=BUSY:20110315T160000/20110315T180000 SUMMARY;LANGUAGE=en-US:The colloq talk by Prof. Howard A. Stone DISTINGUIS HED POLLAK LECTURE about Bacteria, Biofilms and Fluid Dynamics: Elementary Flows and Unexpected Phenomena at 2011-03-15 16:00:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 641 Lady Davis Bld. UID:111ec4410400012720 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110316T113000 DTEND;TZID="Asia/Jerusalem":20110316T133000 DTSTAMP;TZID="Asia/Jerusalem":20110316T113000 FREEBUSY;FBTYPE=BUSY:20110316T113000/20110316T133000 SUMMARY;LANGUAGE=en-US:The phd talk by Ido Ben-Zvi about Causality, Knowle dge and Coordination in Distributed Systems at 2011-03-16 11:30:00 DESCRIPTION;LANGUAGE=en-US:Coordinating the proper ordering of events acro ss remote sites is a central task of distributed applications. In async hronous systems, such coordination depends in an essential way upon mess age chains, as captured by Lamport's happened-before relation. The relat ion provides a useful approximation of causality, in the sense that in a synchronous systems two event can only be causally related if they are L amport related. The talk will consider coordination and causality in synchronous systems, where upper bounds are available on message transm ission times over each channel, and processes share a global clock. Here , active coordination is required whenever a spontaneous external input is meant to trigger an ordered sequence of responses across various site s. We capture the essence of such a coordination task in a proposed clas s of coordination problems called Ordered Response. Within this framewor k we embark on a search for a similar notion of causality. We will not t ouch upon knowledge in any great depth, but consider that in a synchronous setting both message chains and the passage of time can be used to spre ad information across the system, and hence to enable coordination. Indeed, it turns out that the synchronous analog for Lamport's causality is a structure that carefully combines message chains and the existing u pper bounds on transmission times. This causal structure, called the cen tipede, is shown to be necessary in every solution of the Ordered Respon se problem. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Meyer 861 (Electrical Engineering building) UID:111ec4410400012780 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110316T123000 DTEND;TZID="Asia/Jerusalem":20110316T143000 DTSTAMP;TZID="Asia/Jerusalem":20110316T123000 FREEBUSY;FBTYPE=BUSY:20110316T123000/20110316T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Moran Feldman (CS, Techn ion) about Theory Seminar: Submodular Secretary Problems at 2011-03-16 12:30:00 DESCRIPTION;LANGUAGE=en-US:The Classical Secretary Problem was introduced during the 60's of the 20-th century (nobody is sure exactly when). Since its introduction, many variants of the problem have been proposed and rese arched. In the classical secretary problem, and many of its variant, the i nput (which is a set of secretaries, or elements) arrives in a random orde r. We applied to the secretary problem a simple observation which states t hat the random order of the input can be generated by independently choosi ng a random continuous arrival time for each secretary. Surprisingly, this simple observation enables us to improve the competitive ratio of several known and studied variants of the secretary problem. In addition, in some cases the proofs we provide assuming random arrival times are shorter and simpler in comparison to existing proofs. In the talk, I will describe the variants for which we provide the improvements, and will present our a lgorithm for one of these variants. The talk is self contained. No previou s knowledge in submodular optimization or online algorithms is assumed.

Based on a joint work with Seffi Naor and Roy Schwartz. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012800 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110317T143000 DTEND;TZID="Asia/Jerusalem":20110317T163000 DTSTAMP;TZID="Asia/Jerusalem":20110317T143000 FREEBUSY;FBTYPE=BUSY:20110317T143000/20110317T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Prof. Howard A. Stone DISTINGUIS HED POLLAK LECTURE about Unusual Dynamics in Multiphase Flows : (I) Charge d Drops, (II) A Variant on Viscous Fingering, and (III) Shear-Enhanced Dif fusion at 2011-03-17 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012710 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110322T113000 DTEND;TZID="Asia/Jerusalem":20110322T133000 DTSTAMP;TZID="Asia/Jerusalem":20110322T113000 FREEBUSY;FBTYPE=BUSY:20110322T113000/20110322T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Ohad Ben-Shahar (Computer Sc ience, Ben Gurion University) about Pixel Club: Filling in the Gaps the Mind's Way: Curve Completion in the Tangent Bundle at 2011-03-22 11:30:0 0 DESCRIPTION;LANGUAGE=en-US:The phenomenon of visual curve completion, wher e the visual system completes the missing part (e.g., due to occlusion) between two contour fragments, is a major problem in perceptual organi zation research. Previous computational approaches for the shape of the completed curve typically follow formal descriptions of desired, image-b ased perceptual properties (e.g, minimum total curvature, roundedness, etc...). Unfortunately, however, it is difficult to determine such desi red properties psychophysically and indeed there is no consensus in the literature for what they should be. Instead, in this paper we suggest to exploit the fact that curve completion occurs in early vision in order to formalize the problem in a space that explicitly abstracts the primar y visual cortex, namely the unit tangent bundle R2 ×S1. We show that a basic principle of “minimum energy consumption” in this space, i.e ., a minimum length completion, entails desired perceptual properties fo r the completion in the image plane. We present formal theoretical analy sis and numerical solution methods, we show results on natural images an d their advantage over existing popular approaches, and we discuss how our theory explains recent findings from the perceptual literature usin g basic principles only. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012790 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110322T143000 DTEND;TZID="Asia/Jerusalem":20110322T163000 DTSTAMP;TZID="Asia/Jerusalem":20110322T143000 FREEBUSY;FBTYPE=BUSY:20110322T143000/20110322T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Gill Barequet about Formulae and Growth Rates of High-Dimensional Polycubes at 2011-03-22 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012590 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110323T123000 DTEND;TZID="Asia/Jerusalem":20110323T143000 DTSTAMP;TZID="Asia/Jerusalem":20110323T123000 FREEBUSY;FBTYPE=BUSY:20110323T123000/20110323T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Sigurd Torkel Meldgaard (University of Aarhus, Denemark) about Theory Seminar: Oblivious RAM witho ut Random Oracles at 2011-03-23 12:30:00 DESCRIPTION;LANGUAGE=en-US:We present an algorithm for implementing a secu re oblivious RAM where the access pattern is perfectly hidden in the infor mation theoretic sense, without assuming that the CPU has access to a rand om oracle. In addition we prove a lower bound on the amount of randomness needed for implementing an information theoretically secure oblivious RAM.

Authors: Ivan Damgård, Sigurd Meldgaard, Jesper Buus Niels en ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012820 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110324T103000 DTEND;TZID="Asia/Jerusalem":20110324T123000 DTSTAMP;TZID="Asia/Jerusalem":20110324T103000 FREEBUSY;FBTYPE=BUSY:20110324T103000/20110324T123000 SUMMARY;LANGUAGE=en-US:The cggc talk by Michael Floater (University of Osl o, Norway) about CGGC Seminar: Barycentric Interpolation and Mappings on S mooth Convex Domains at 2011-03-24 10:30:00 DESCRIPTION;LANGUAGE=en-US:In a recent paper, Warren, Schaefer, Hirani, an d Desbrun proposed a simple method of interpolating a function define d on the boundary of a smooth convex domain, using an integral kernel with properties similar to those of barycentric coordinates on simplexes.

When applied to vector-valued data, the interpolation can ma p one convex region into another, with various potential applications in computer graphics, such as curve and image deformation. In this paper we establish some basic mathematical properties of barycentric kerne ls in general, including the interpolation property and a formula for the Jacobian of the mappings they generate.

We then use this formula to prove the injectivity of the mapping of Warren et al. Thi s is joint work with Jiri Kosinka. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012830 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110328T183000 DTEND;TZID="Asia/Jerusalem":20110328T203000 DTSTAMP;TZID="Asia/Jerusalem":20110328T183000 FREEBUSY;FBTYPE=BUSY:20110328T183000/20110328T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Guy Keren about Hai fux Club: The Story of Alice and Bob - the I/O Requests (Part II) at 2011- 03-28 18:30:00 DESCRIPTION;LANGUAGE=en-US:In this story, we'll follow the life story of a lice - a file-systemized I/O request, and bob - a raw-device I/O request, from their birth, until they reach heaven (the disk or the USB camera or.. .). In addition, we will cover some system parameters that affect I/O requ ests, the buffer cache, disk I/O schedulers and tools used to track and co unt I/O (including - why is process-based I/O accounting so tricky). ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400012760 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110329T113000 DTEND;TZID="Asia/Jerusalem":20110329T133000 DTSTAMP;TZID="Asia/Jerusalem":20110329T113000 FREEBUSY;FBTYPE=BUSY:20110329T113000/20110329T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Albert Achtenberg (EE, Techn ion) about Pixel Club: Unmixing of Images Mixed by Position Varying Media at 2011-03-29 11:30:00 DESCRIPTION;LANGUAGE=en-US:We address the open problem of blindly separati ng single-path position varying image mixtures, without having prior inf ormation about the sources. We assume that the mixing system's spatial d istortion and attenuation change with position. A staged method for esti mating the mixing models is used in turn to recover source signals from such mixtures. Our method is based on a Staged Sprase Component Analysis (SSCA) of the mixtures. Our method consists of three stages: aligning the signals to estimate the spatial distortion component of the mixing system; classifying the sparse signal samples to their estimated sources and estimating the spatial attenuation component of the mixing system; and finally inverting the mixing system to recover the sources. Small er ror in the spatial distortion component leads to a significant degradati on in separation quality. However, in practice, uncertainty is associate d with the model estimation stage, due to: noisy samples; bad spatial sp read of samples; mixed samples in the sparse representation and more. We propose a solution by adding a step of model refinement that is based o n simple image quality measures that would allow us to improve separatio n results. We test some standard methods and propose a new aproach based on Phase Congruency measure. M.Sc. thesis under the supervision of Prof . Zeevi ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400012870 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110330T113000 DTEND;TZID="Asia/Jerusalem":20110330T133000 DTSTAMP;TZID="Asia/Jerusalem":20110330T113000 FREEBUSY;FBTYPE=BUSY:20110330T113000/20110330T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by David Carmel and Dafna Sheinwald (IBM Research, Haifa) about ceClub: IBM Watson and the Jeopardy! Challen ge at 2011-03-30 11:30:00 DESCRIPTION;LANGUAGE=en-US:Over the last days, millions of viewers witness ed computing history being made as IBM's Watson question answering system defeated Jeopardy! quiz show champions Brad Rutter and Ken Jennings. Watso n is an application of advanced natural language processing, information r etrieval, knowledge representation and reasoning, and machine learning tec hnologies to the field of open domain question answering. Watson runs on a cluster of 90 IBM Power 750 servers in 10 racks with a total of 2880 POWE R7 processor cores and 16 Terabytes of RAM, enabling it to respond within less than 3 seconds. The deep analytic capabilities and systems technologi es that enabled Watson to win at Jeopardy! now can be customized for speci fic industry lexicons and applications to help humans make smarter, faster and more effective decisions.

In this talk we will tell mor e about the challenge, Watson's architecture and technologies, our contrib utions to the system, and some insights from the man vs. machine match. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 3 UID:111ec4410400012860 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110330T123000 DTEND;TZID="Asia/Jerusalem":20110330T143000 DTSTAMP;TZID="Asia/Jerusalem":20110330T123000 FREEBUSY;FBTYPE=BUSY:20110330T123000/20110330T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Amir Yehudaioff (Mathema tics, Technion) about Theory Seminar: On the Rank of Design Matrices with Applications at 2011-03-30 12:30:00 DESCRIPTION;LANGUAGE=en-US:A design matrix is a matrix whose attern of zeo s/nonzeros satisfies a certain design-like condition. We will first prove that the rank of any design matrix is high.

We shall discuss two applications of this rank lower bound: (1) Impossibility results for 2-query locally correctable codes over real/complex numbers, and (2) ge neralization of results in combinatorial geometry, for example, a robust analog of the Sylvester-Gallai theorem.

Joint work with Boaz Barak, Zeev Dvir and Avi Wigderson. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012850 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110405T113000 DTEND;TZID="Asia/Jerusalem":20110405T133000 DTSTAMP;TZID="Asia/Jerusalem":20110405T113000 FREEBUSY;FBTYPE=BUSY:20110405T113000/20110405T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Yacov Hel-Or (CS, The Interd isciplinary Center) about Pixel Club: Pattern Matching under Non Linear To ne-Mapping at 2011-04-05 11:30:00 DESCRIPTION;LANGUAGE=en-US:A fast pattern matching scheme termed Matching by Tone Mapping (MTM) is introduced which allows matching under non-line ar tone mappings. We exploit the recently introduced Slice Transform to implement a fast computational scheme requiring computational time simil ar to the fast implementation of Normalized Cross Correlation (NCC). In fact, the MTM measure can be viewed as a generalization of the NCC for n on-linear mappings and actually reduces to NCC when mappings are restric ted to be linear. The MTM is shown to be invariant to non-linear tone ma ppings, and is empirically shown to be highly discriminative and robust to noise. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012920 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110405T143000 DTEND;TZID="Asia/Jerusalem":20110405T163000 DTSTAMP;TZID="Asia/Jerusalem":20110405T143000 FREEBUSY;FBTYPE=BUSY:20110405T143000/20110405T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Uri Zwick about Subexponential l ower bounds for randomized pivoting rules for the simplex algorithm at 201 1-04-05 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012840 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110406T113000 DTEND;TZID="Asia/Jerusalem":20110406T133000 DTSTAMP;TZID="Asia/Jerusalem":20110406T113000 FREEBUSY;FBTYPE=BUSY:20110406T113000/20110406T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Eitan Zahavi (Mellanox Technolog ies) about ceClub: Fat-Trees Routing and Node Ordering Providing Contentio n Free Traffic for MPI Global Collectives at 2011-04-06 11: 30:00 DESCRIPTION;LANGUAGE=en-US:As the size of High Performance Computing clust ers grows, the increasing probability of interconnect hot spots degrades t he latency and effective bandwidth the network provides. This paper presen ts a solution to this scalability problem for real life constant bisection al-bandwidth fat-tree topologies. It is shown that maximal bandwidth and c ut-through latency can be achieved for MPI global collective traffic. To f orm such congestion-free configuration, MPI programs should utilize collec tive communication, MPI-node-order should be topology aware, and the packe ts routing should match the MPI communication patterns. First, we show tha t MPI collectives can be classified into unidirectional and bidirectional shifts. Using this property, we propose a scheme for congestion-free routi ng of the global collectives in fully and partially populated fat trees ru nning a single job. Simulation results of the proposed routing, MPI-node-o rder and communication patterns show a 40% throughput improvement over pre viously published results for all-to-all collectives. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 861 UID:111ec4410400012960 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110406T123000 DTEND;TZID="Asia/Jerusalem":20110406T143000 DTSTAMP;TZID="Asia/Jerusalem":20110406T123000 FREEBUSY;FBTYPE=BUSY:20110406T123000/20110406T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Roy Schwartz (CS, Techni on) about Theory Seminar: Min-Max Graph Partitioning and Small Set Expansi on at 2011-04-06 12:30:00 DESCRIPTION;LANGUAGE=en-US:We study graph partitioning problems from a min -max perspective, in which an input graph on $n$ vertices should be part itioned into $k$ parts, and the objective is to minimize the maximum num ber of edges leaving a single part. The two main versions we consider ar e where the $k$ parts need to be of equal-size, and where they must sepa rate a set of $k$ given terminals. We consider a common generalization o f these two problems, and design for it an $O(\sqrt{\log n}\log^{3/2} k) $-approximation algorithm. This improves over an $O(\log^2 n)$ approxima tion for the version of separating $k$ given terminals due to Svitkina a nd Tardos \cite{ST04}, and roughly $O(k\log n)$ approximation for the ve rsion where the parts need to be of equal size that follows from other p revious work.

The main tool we use is a new approximation alg orithm for \emph{$\rho$-unbalanced-cut}, the problem of finding in an in put graph $G=(V,E)$ a set $S\subseteq V$ of size $|S|=\rho n$ that minim izes the number of edges leaving $S$. We provide a bicriteria approximat ion of $O(\sqrt{\log{n}\log{(1/\rho)}})$ for this problem. Note that the special case $\rho = 1/2$ is just the \emph{minimum bisection} problem, and indeed our bound generalizes that of Arora, Rao and Vazirani \cite{ ARV08} which only applies to the case where $\rho = \Omega(1)$. Our algo rithm also works for the closely related \emph{small-set-expansion} prob lem, which asks for a set $S\subseteq V$ of size $|S| \leq \rho n$ with minimum conductance (edge-expansion), and was suggested recently by Ragh avendra and Steurer~\cite{RS10}. In fact, our algorithm handles more gen eral, weighted, versions of both problems. Previously, an $O(\log n)$ tr ue approximation for both $\rho$-unbalanced-cut and small-set-expansion was known following from R\"acke~\cite{Racke08}.

Joint work w ith: Nikhil Bansal, Robi Krauthgamer, Konstantin Makarychev, Viswanath N agarajan and Seffi Naor. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 401 (Note unusual room) UID:111ec4410400012970 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110406T130000 DTEND;TZID="Asia/Jerusalem":20110406T150000 DTSTAMP;TZID="Asia/Jerusalem":20110406T130000 FREEBUSY;FBTYPE=BUSY:20110406T130000/20110406T150000 SUMMARY;LANGUAGE=en-US:The msc talk by Roi Adadi about Improved genome-sca le metabolic modeling utilizing enzyme kinetic parameters at 2011-04-06 13 :00:00 DESCRIPTION;LANGUAGE=en-US:Genome-scale metabolic models enable to success fully predict a variety of metabolic phenotype in microorganisms. Still, t he integration of metabolic networks with various 'omics' data towards the prediction of metabolic flux remains an open challenge. Here, we show tha t enzyme kinetic parameters are significantly correlated with measured flu xes in E. coli under various conditions, providing a higher correlation th an that achieved by measured gene expression data. Based on the latter, we developed a novel constraint-based modeling method, Enzyme Solvent-Capaci ty Flux Balance Analysis (ESC-FBA), which predicts cellular metabolic stat e by utilizing prior data on enzyme turn-over rates and molecular weights. In contrast to previous attempts to predict global metabolic flux by util izing kinetic constants, our approach thoroughly accounts for enzymatic re quirements for catalyzing metabolic flux, considering isozymes, protein co mplexes, and multi-functional enzymes. ESC-FBA is shown to significantly i mprove the prediction accuracy of various metabolic phenotypes in E. coli, including growth rate, flux rates, enzyme concentrations and gene express ion levels, in comparison with state-of-the art computational approaches. A specifically interesting demonstration of ESC-FBA's applicability involv es the prediction of lag-phase length when E. coli is switched between gro wth media, shedding light on this intriguing biological phenomenon. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012890 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110406T143000 DTEND;TZID="Asia/Jerusalem":20110406T163000 DTSTAMP;TZID="Asia/Jerusalem":20110406T143000 FREEBUSY;FBTYPE=BUSY:20110406T143000/20110406T163000 SUMMARY;LANGUAGE=en-US:The msc talk by Avishay Tal about On the Minimal Fo urier Degree of Symmetric Boolean Functions at 2011-04-06 14:30:00 DESCRIPTION;LANGUAGE=en-US:It is well-known that any Boolean function f:{- 1,+1}^n \to {-1,+1} can be written uniquely as a polynomial f(x) = \sum_{S subset [n]} f_s \prod_{i in S} x_i. The collection of coefficients (f_S's ) this expression are referred to (with good reason) as the Fourier spectr um of f. The Fourier spectrum has played a central role in modern computer science by converting combinatorial and algorithmic questions about f int o algebraic or analytic questions about the spectrum. In this talk I will focus on a basic feature of the Fourier spectrum, namely the minimal Fourier degree, or the size of the smallest non-empty set S such that f_S is non-zero. For every symmetric function *except the parity function* we show that the minimal Fourier degree is at most O(Gamma(n)) where Gamma(m ) < m^{0.525} is the largest gap between consecutive prime numbers in {1,. ..,m}. This improves the previous result of Kolountzakis et al. (Combinato rica '09) who showed that the minimal Fourier degree is at most k/log k. A s an application we obtain a new analysis of the PAC learning algorithm fo r symmetric juntas, under the uniform distribution, of Mossel et al. (STOC '03). ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012880 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110407T150000 DTEND;TZID="Asia/Jerusalem":20110407T170000 DTSTAMP;TZID="Asia/Jerusalem":20110407T150000 FREEBUSY;FBTYPE=BUSY:20110407T150000/20110407T170000 SUMMARY;LANGUAGE=en-US:The colloq talk by David Woodruff SPECIAL GUEST TAL K note unusual hour about Near-Optimal Private Approximation Protocols via a Black Box Transformation at 2011-04-07 15:00:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012950 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110410T113000 DTEND;TZID="Asia/Jerusalem":20110410T133000 DTSTAMP;TZID="Asia/Jerusalem":20110410T113000 FREEBUSY;FBTYPE=BUSY:20110410T113000/20110410T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Ariel Tankus (Biomedical Eng ineering, Technion) about Pixel Club: Decoding Neural Patterns for Brain-C omputer Interfaces at 2011-04-10 11:30:00 DESCRIPTION;LANGUAGE=en-US:Brain-computer interfaces (BCI) are devices tha t can decode physiological signals from the brain and convert them into ac tions in a manner that reflects the brain’s intention. Their goal is to replace or restore lost function in paralyzed humans by routing movement- relate​d signals from the brain, around damaged parts of the nervous sys tem, to external effectors. My research is aimed at developing a new gene ration of brain-computer interfaces at the single cell level with human pa rticipants. One type of BCIs I currently develop is geared towards direct brain control over object movement, based directly on the goal movement t hat the brain desires for that object, as distinguished from hand movement s that the brain may plan for manipulating the object to achieve this move ment. Thus, objects will be moved in the manner the individual desires, b ut without decoding any hand movement.

The other type of B CI in the focus of my talk is aimed to decode natural speech production fr om brain activity. Recent data I have collected during speech production and imagery of speech production demonstrate robust differential activatio n during utterance of different phonemes in single neurons from the entorh inal cortex, anterior cingulate gyrus and superior temporal gyrus. We obt ain high accuracy in decoding the phonemes from the neural pattern. Decod ing these activations, even with a minimal set of words or phonemes, bears an enormous promise for locked-in patients following, for example, amyotr ophic lateral sclerosis (ALS) or stroke, to be able to “speak” again. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012990 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110410T123000 DTEND;TZID="Asia/Jerusalem":20110410T143000 DTSTAMP;TZID="Asia/Jerusalem":20110410T123000 FREEBUSY;FBTYPE=BUSY:20110410T123000/20110410T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Ilan Nehama (Hebrew Univ ersity of Jerusalem) about Theory Seminar: Approximate Judgement Aggregati on at 2011-04-10 12:30:00 DESCRIPTION;LANGUAGE=en-US:In this work we analyze judgement aggregation p roblems in which a group of agents independently votes on a set of complex propositions that has some interdependency constraint between them (e.g., transitivity when describing preferences). We generalize the previous res ults by studying approximate judgement aggregation. We relax the main two constraints assumed in the current literature, Consistency and Independenc e and consider mechanisms that only approximately satisfy these constraint s, that is, satisfy them up to a small portion of the inputs. The main que stion we raise is whether the relaxation of these notions significantly al ters the class of satisfying aggregation mechanisms. The recent works for preference aggregation of Kalai, Mossel, and Keller fit into this framewor k. The main result of this paper is that, as in the case of preference agg regation, in the case of a subclass of a natural class of aggregation prob lems termed `truth-functional agendas', the set of satisfying aggregation mechanisms does not extend non-trivially when relaxing the constraints.

In the talk I plan to describe this framework of judgement aggr egation, describe the question of approximate aggregation and show resembl ance to current works in the field of local testing. In addition, I will d escribe few agendas in which we can characterize the approximate aggregati on mechanisms. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013010 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110411T183000 DTEND;TZID="Asia/Jerusalem":20110411T203000 DTSTAMP;TZID="Asia/Jerusalem":20110411T183000 FREEBUSY;FBTYPE=BUSY:20110411T183000/20110411T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Tomer Ashur about H aifux Club: How to Spread Knowledge Throughout the World Wikipedia While W earing Only Your Slippers (or Wikmedia, and free content projects) at 20 11-04-11 18:30:00 DESCRIPTION;LANGUAGE=en-US:Since its first emergence in 2001, Wikipedia ha d grown drastically to become the fifth most viewed website over the Inter net in 2012. With over 12,000,000 articles in more than 250 languages this is now the largest source of information ever existed. The Wikimedia foun dation has been founded in 2003 as a non-profit organization to support Wi kipedia as well as other online and offline free content projects.

In this talk I will present Wikipedia and other projects of the Wik imedia foundation and how they work. I will show how one can become a Wiki pedia editor and why would she want to do so. We will talk about readers, editors, Neutrality, reliability, dispute resolution, style and free conte nt. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400012900 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110413T113000 DTEND;TZID="Asia/Jerusalem":20110413T133000 DTSTAMP;TZID="Asia/Jerusalem":20110413T113000 FREEBUSY;FBTYPE=BUSY:20110413T113000/20110413T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Eran Yahav (CS, Technion) about ceClub: Abstraction-Guided Synthesis of Synchronization at 2011-04-13 1 1:30:00 DESCRIPTION;LANGUAGE=en-US:In this talk I will present a framework for syn thesizing efficient synchronization in concurrent programs, a task known t o be difficult and error-prone when done manually. The framework is based on abstract interpretation and can infer synchronization for infinite stat e programs. Given a program, a specification, and an abstraction, we infer synchronization that avoids all (abstract) interleavings that may violate the specification, but permits as many valid interleavings as possible. I will show application of this general idea for automatic inference of ato mic sections and memory fences in programs running over relaxed memory mod els.

Joint work with Michael Kuperstein, Martin Vechev, and Greta Yorsh. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 3 UID:111ec4410400013000 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110413T133000 DTEND;TZID="Asia/Jerusalem":20110413T153000 DTSTAMP;TZID="Asia/Jerusalem":20110413T133000 FREEBUSY;FBTYPE=BUSY:20110413T133000/20110413T153000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Martin Akerman (Ad rian Krainer Lab / Michael Zhang Lab, Cold Spring Harbor Laboratory) about Bioinformatics Forum: Splicing the Wires: Finding Connections between Bio logical Networks and the Core **Spliceosome** at 2011-04-13 13:30:00 DESCRIPTION;LANGUAGE=en-US:The major spliceosome is a multi-component and highly dynamic complex that carries out the tightly regulated steps of s plicing. It is composed of hundreds of proteins, including five small nu clear ribonucleoprotein complexes (snRNPs) that catalyze>99% of all pre- mRNA splicing events. In addition, there are alternative splicing regula tors, such as the SR and hnRNP proteins, that either activate or repress a subset of the splicing events. Members of these protein families are known to interact with components of the spliceosome; however, which int eractions are direct and functional, and whether they participate in spl iceosome assembly is unknown. I will present a high-throughput proteomic approach to assemble and explore the interactome network of four well k nown alternative splicing factors: SRSF1 (SF2/ASF), SRSF6 (SRp55), hnRNP A1 and the brain/muscle specific splicing factor RBFOX1. Using a combina tion of affinity purification and mass spectrometry, we generated protei n-protein interaction (PPI) networks of each splicing factor. The analys is of the reconstructed networks revealed specific points of interaction between these splicing factors and the spliceosome, as well as difference s in their global connectivity. Unexpectedly, we also observed links bet ween splicing factors and other biological networks, as in the case of R BFOX1 and a complex of proteins involved in Alzheimer. Another useful wa y to address the relationship of splicing to other biological networks w ould be through the study of Protein-RNA interactions, between specific splicing factors and their alternative splicing targets. It is possible to uncover such connections using high throughput technologies like RNA deep sequencing. I will present SpliceTrap, our newly developed method t o quantify exon inclusion levels using paired-end RNA-seq data. Unlike o ther tools, which focus on the assembly of full-length transcript isofor ms, SpliceTrap approaches the expression level estimation of each exon a s an independent problem. In addition, SpliceTrap can identify alternati ve splicing events under a single cellular condition, without requiring a background set of reads to estimate relative splicing changes. I will show a case of study in which we applied SpliceTrap to uncover splicing targets of SRSF1. In summary, through the combination of high throughput proteomics and transcriptomics, we are aiming to uncover connections th at would help us to understand how the process of splicing is integrated in the global network of the cell. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400012930 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110413T140000 DTEND;TZID="Asia/Jerusalem":20110413T160000 DTSTAMP;TZID="Asia/Jerusalem":20110413T140000 FREEBUSY;FBTYPE=BUSY:20110413T140000/20110413T160000 SUMMARY;LANGUAGE=en-US:The msc talk by Israel Shalom about Online Universa l Facility Location at 2011-04-13 14:00:00 DESCRIPTION;LANGUAGE=en-US:Facility location problems concern assigning re quests to servers. The goal is to minimize the total cost which consists o f the moving costs and the latency costs. The moving costs depend on the d istances between matcheded request/server pairs, while the latency costs d epend on the number of requests matched to each server. Facility location problems arise in many natural settings of resource sharing, such as serve r assignment in cloud computing, network routing and more. Universal Facil ity Location (UniFL) is the broadest framework for such problems, and it a lso generalizes load balancing and bipartite matching problems. This t alk is about the online version of UniFL, and is based on the work done wi th and under the supervision of Prof. Seffi Naor. We propose a greedy algo rithm and consider various cases based on different latency (congestion) f unctions. Through somewhat intricate analysis, we find that for many cases the greedy algorithm provides a constant competitive-ratio, and there are some settings in which no algorithm can offer *any* competitive-ratio. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012980 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110426T143000 DTEND;TZID="Asia/Jerusalem":20110426T163000 DTSTAMP;TZID="Asia/Jerusalem":20110426T143000 FREEBUSY;FBTYPE=BUSY:20110426T143000/20110426T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Manor Mendel about Ramsey-type t heorems for metric spaces at 2011-04-26 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013020 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110427T113000 DTEND;TZID="Asia/Jerusalem":20110427T133000 DTSTAMP;TZID="Asia/Jerusalem":20110427T113000 FREEBUSY;FBTYPE=BUSY:20110427T113000/20110427T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Benny Pinkas (CS, Bar-Ilan Unive rsity) about ceClub: Side channels in Cloud Services: The Case of Deduplic ation in Cloud Storage at 2011-04-27 11:30:00 DESCRIPTION;LANGUAGE=en-US:The talk will discuss deduplication, a form of compression in which duplicate copies of files are replaced by links to a single copy. Deduplication is known to reduce the space and bandwidth requ irements of Cloud storage services by more than 90%, and is most effective when applied across multiple users.

We study the privacy im plications of cross-user deduplication. We demonstrate how deduplication c an be used as a side channel which reveals information about the contents of files of other users, as a covert channel by which malicious software c an communicate with its control center, or as a method to retrieve files a bout which you have only partial information.

Due to the hig h savings offered by cross-user deduplication, cloud storage providers are unlikely to stop ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 3 UID:111ec4410400013040 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110427T123000 DTEND;TZID="Asia/Jerusalem":20110427T143000 DTSTAMP;TZID="Asia/Jerusalem":20110427T123000 FREEBUSY;FBTYPE=BUSY:20110427T123000/20110427T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Ilan Orlov (Bar-Ilan Uni versity) about Theory Seminar: Protocols for Multiparty Coin Toss With Dis honest Majority at 2011-04-27 12:30:00 DESCRIPTION;LANGUAGE=en-US:Generating random bits is a fundamental problem in cryptography. Coin-tossing protocols, which generate a random bit with uniform distribution, are used as a building box in many cryptographic pr otocols. Cleve [STOC 1986] has shown that if at least half of the parties can be malicious, then, in any $r$-round coin-tossing protocol, the malici ous parties can cause a bias of $\Omega(1/r)$ in the bit that the honest p arties output. However, for more than two decades the best known protocols had bias $t/\sqrt{r}$, where $t$ is the number of corrupted parties. Rece ntly, in a surprising result, Moran, Naor, and Segev [TCC 2009] have shown that there is an $r$-round two-party coin-tossing protocol with the optim al bias of $O(1/r)$. We extend Moran et al.~results to the multiparty mode l when less than 2/3 of the parties are malicious. The bias of our protoco l is proportional to $1/r$ and depends on the gap between the number of ma licious parties and the number of honest parties in the protocol. Specific ally, for a constant number of parties or when the number of malicious par ties is somewhat larger than half, we present an $r$-round $m$-party coin- tossing protocol with optimal bias of $O(1/r)$. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013050 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110501T133000 DTEND;TZID="Asia/Jerusalem":20110501T153000 DTSTAMP;TZID="Asia/Jerusalem":20110501T133000 FREEBUSY;FBTYPE=BUSY:20110501T133000/20110501T153000 SUMMARY;LANGUAGE=en-US:The msc talk by Naama Kraus about Context-Sensitive Query Auto-Completion at 2011-05-01 13:30:00 DESCRIPTION;LANGUAGE=en-US:Query auto completion is known to provide poor predictions of the user's query when her input prefix is very short (e.g. , one or two characters). In this work we show that context, such as the user's recent queries, can be used to improve the prediction quality con siderably even for such short prefixes. We propose a context-sensitive qu ery auto completion algorithm, NearestCompletion, which outputs the compl etions of the user's input that are most similar to the context queries. To measure similarity, we represent queries and contexts as high-dimensi onal term-weighted vectors and resort to cosine similarity. The mapping f rom queries to vectors is done through a new query expansion technique th at we introduce, which expands a query by traversing the query recommend ation tree rooted at the query. In order to evaluate our approach, we p erformed extensive experimentation over the public AOL query log. We dem onstrate that when the recent user's queries are relevant to the current query she is typing, then after typing a single character, NearestComplet ion's MRR is 48% higher relative to the MRR of the standard MostPopularCo mpletion algorithm on average. When the context is irrelevant, however, NearestCompletion's MRR is essentially zero. To mitigate this problem, we propose HybridCompletion, which is a hybrid of NearestCompletion with Mo stPopularCompletion. HybridCompletion is shown to dominate both NearestCo mpletion and MostPopularCompletion, achieving a total improvement of 31. 5% in MRR relative to MostPopular- Completion on average. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012810 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110502T183000 DTEND;TZID="Asia/Jerusalem":20110502T203000 DTSTAMP;TZID="Asia/Jerusalem":20110502T183000 FREEBUSY;FBTYPE=BUSY:20110502T183000/20110502T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Nadav Amit (CS, Tec hnion) about Haifux Club: vIOMMU: Efficient IOMMU Emulation at 2011-05-02 18:30:00 DESCRIPTION;LANGUAGE=en-US:Direct device assignment, where a guest virtual machine directlyinteracts with an I/O device without host intervention, i s appealing,because it allows an unmodified (non-hypervisor-aware) guest t oachieve near-native performance. But device assignment for unmodifiedgues ts suffers from two serious deficiencies: (1) it requires pinningof all th e guest's pages, thereby disallowing memory overcommitment,and (2) it expo ses the guest's memory to buggy device drivers.

We solve the se problems by designing, implementing, and exposing anemulated IOMMU (vIO MMU) to the unmodified guest. We employ two noveloptimizations to make vIO MMU perform well: (1) waiting a fewmilliseconds before tearing down an IOM MU mapping in the hope it willbe immediately reused (``optimistic teardown ''), and (2) running the vIOMMU on a sidecore, and thereby enabling for th e first time the useof a sidecore by unmodified guests. Both optimizations are highlyeffective in isolation. The former allows bare-metal to achieve 100%of a 10Gbps line rate. The combination of the two allows an unmodifie dguest to do the same. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400013100 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110503T113000 DTEND;TZID="Asia/Jerusalem":20110503T133000 DTSTAMP;TZID="Asia/Jerusalem":20110503T113000 FREEBUSY;FBTYPE=BUSY:20110503T113000/20110503T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Jonathan Pokrass (Tel Aviv U niversity) about Pixel Club: Correspondence-less Approach to Matching of Deformable Shapes at 2011-05-03 11:30:00 DESCRIPTION;LANGUAGE=en-US:Finding a match between partially available def ormable shapes is a challenging problem with numerous applications. The problem is usually approached by computing local descriptors on a pair o f shapes and then establishing a point-wise correspondence between the t wo. We introduce an alternative correspondence-less approach to matching fragments to an entire shape undergoing a non-rigid deformation. We use diffusion geometric descriptors and optimize over the integration domai ns on which the integral descriptors of the two parts match. The problem is regularized using the Mumford-Shah functional. We show an efficient discretization based on the Ambrosio-Tortorelli approximation generalize d to triangular meshes. Experiments demonstrating the success of the p roposed method are presented.

This work was done under the su pervision of Dr. Alex Bronstein. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013090 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110503T143000 DTEND;TZID="Asia/Jerusalem":20110503T163000 DTSTAMP;TZID="Asia/Jerusalem":20110503T143000 FREEBUSY;FBTYPE=BUSY:20110503T143000/20110503T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Liam Roditty about The Minimum W eight Cycle Problem at 2011-05-03 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400012910 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110504T123000 DTEND;TZID="Asia/Jerusalem":20110504T143000 DTSTAMP;TZID="Asia/Jerusalem":20110504T123000 FREEBUSY;FBTYPE=BUSY:20110504T123000/20110504T143000 SUMMARY;LANGUAGE=en-US:The phd talk by Ilya Volkovich about Black-Box Iden tity Testing of Depth-4 Multilinear Circuits at 2011-05-04 12:30:00 DESCRIPTION;LANGUAGE=en-US:A central problem in algebraic complexity theor y and algorithms design is the problem of Polynomial Identity Testing (PI T): given an arithmetic circuit $C$ over a field $\F$, with input variable s $x_1, x_2, ... , x_n$, determine whether $C$ computes the identically ze ro polynomial. Numerous applications and connections to other algorithmic and number theoretic problems further emphasize the significance of PIT. Among the examples are algorithms for finding perfect matchings in graphs \cite{Lovasz79,MVV87}, primality testing \cite{AKS04}, and many more. We study the problem of PIT for multilinear $\Sigma \Pi \Sigma \Pi (k) $ circuits, i.e. multilinear depth-4 circuits with fan-in k at the top + ga te. We give the first polynomial-time deterministic PIT algorithm for such circuits. Our results also hold in the black-box setting. The importance of this model arises from \cite{AgrawalVinay08}, where it was shown tha t derandomizing black-box polynomial identity testing for general depth-4 circuits implies a derandomization of polynomial identity testing (PIT) fo r general arithmetic circuits. We obtain our results by showing a strong structural result for multilinear $\Sigma \Pi \Sigma \Pi (k) $ circuits that compute the zero polynomial. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400012040 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110504T123000 DTEND;TZID="Asia/Jerusalem":20110504T143000 DTSTAMP;TZID="Asia/Jerusalem":20110504T123000 FREEBUSY;FBTYPE=BUSY:20110504T123000/20110504T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Ilya Volkovich (CS Techn ion) about Theory Seminar: Black-Box Identity Testing of Depth-4 Multilin ear Circuits at 2011-05-04 12:30:00 DESCRIPTION;LANGUAGE=en-US:A central problem in algebraic complexity theor y and algorithms design is the problem of Polynomial Identity Testing (P IT): given an arithmetic circuit $C$ over a field $\F$, with input variabl es $x_1, x_2, ... , x_n$, determine whether $C$ computes the identically z ero polynomial. Numerous applications and connections to other algorithmi c and number theoretic problems further emphasize the significance of PIT. Among the examples are algorithms for finding perfect matchings in graphs \cite{Lovasz79,MVV87}, primality testing \cite{AKS04}, and many more. We study the problem of PIT for multilinear $\Sigma \Pi \Sigma \Pi (k) $ circuits, i.e. multilinear depth-4 circuits with fan-in k at the top + gate. We give the first polynomial-time deterministic PIT algorithm for s uch circuits. Our results also hold in the black-box setting. The importa nce of this model arises from \cite{AgrawalVinay08}, where it was shown that derandomizing black-box polynomial identity testing for general dept h-4 circuits implies a derandomization of polynomial identity testing (PIT ) for general arithmetic circuits. We obtain our results by showing a st rong structural result for multilinear $\Sigma \Pi \Sigma \Pi (k) $ cir cuits that compute the zero polynomial. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013070 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110504T123000 DTEND;TZID="Asia/Jerusalem":20110504T143000 DTSTAMP;TZID="Asia/Jerusalem":20110504T123000 FREEBUSY;FBTYPE=BUSY:20110504T123000/20110504T143000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Yossi Matias (Head, Israel R&D C enter, Google) about ceClub: Search Flavors - Trends and Opportunities at 2011-05-04 12:30:00 DESCRIPTION;LANGUAGE=en-US:This talk will discuss some recent developments in search, emerging in various shapes and forms. We will highlight some c hallenges, and point to some search trends that play an increasing role in multiple domains. We will also discuss the power of data and the signific ant role of cloud technologies in facilitation of new opportunities. Some of the core technologies and global innovations developed in Google's R&D center in Israel will be highlighted. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Meyer 1003 UID:111ec4410400013080 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110505T143000 DTEND;TZID="Asia/Jerusalem":20110505T163000 DTSTAMP;TZID="Asia/Jerusalem":20110505T143000 FREEBUSY;FBTYPE=BUSY:20110505T143000/20110505T163000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Yaron Singer (CS, UC Berkeley) a bout ceClub: How to win Friends and Influence People, Truthfully at 20 11-05-05 14:30:00 DESCRIPTION;LANGUAGE=en-US:Throughout the past decade there has been exten sive research on algorithmic and data mining techniques for solving the problem of influence maximization in social networks: if one can convinc e a subset of individuals to influence their friends to adopt a new prod uct or technology, which subset should be selected so that the spread of influence in the social network is maximized?

Despite the pr ogress in modeling and techniques, the incomplete information aspect of problem has been largely overlooked. While the network structure is ofte n available, the inherent cost individuals have for influencing friends is difficult to extract.

In this talk we will discuss mechani sms that extract individuals' costs in well studied models of social net work influence. We follow the mechanism design framework which advocates for truthful mechanisms that use allocation and payment schemes that in centivize individuals to report their true information. Beyond their pro vable theoretical guarantees, the mechanisms work well in practice. To s how this we will use results from experiments performed on the mechanica l turk platform and social network data that provide experimental eviden ce of the mechanisms' effectiveness. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013140 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110505T143000 DTEND;TZID="Asia/Jerusalem":20110505T163000 DTSTAMP;TZID="Asia/Jerusalem":20110505T143000 FREEBUSY;FBTYPE=BUSY:20110505T143000/20110505T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Yaron Singer SPECIAL GUEST LECTU RE about How to win Friends and Influence People, Truthfully at 2011-05-05 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013150 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110511T113000 DTEND;TZID="Asia/Jerusalem":20110511T133000 DTSTAMP;TZID="Asia/Jerusalem":20110511T113000 FREEBUSY;FBTYPE=BUSY:20110511T113000/20110511T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Roman Manevich (UT Austin) about ceClub: A Shape Analysis for Optimizing Parallel Graph Programs at 2011-0 5-11 11:30:00 DESCRIPTION;LANGUAGE=en-US:In the first part of the talk I will give a hig h-level introduction to Galois, a framework for designing and implementing parallel graph algorithms and related concepts. I will explain what are o rdered/unordered graph algorithms and how optimistic parallelism is achiev ed using transactional boosting.

In the second part of the t alk I will describe a new shape analysis, which is used for analyzing grap h algorithms written with Galois. The shape analysis infers properties tha t can be used to optimize the graph algorithm by reducing two of the main overheads of the system --- synchronization overheads and rollback logging overheads, We implemented the analysis on top of TVLA and applied it to s everal graph algorithms. The optimized (parallel) graph algorithms run 2x to 12x faster than the unoptimized algorithms.

This talk is based on joint work with Dimitrios Prountzos, Keshav Pingali, and Kathryn McKinley, presented at POPL 2011. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 3 UID:111ec4410400013180 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110511T123000 DTEND;TZID="Asia/Jerusalem":20110511T143000 DTSTAMP;TZID="Asia/Jerusalem":20110511T123000 FREEBUSY;FBTYPE=BUSY:20110511T123000/20110511T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Rahul Santhanam (Edinbur gh University) about Theory Seminar: Using Random Restrictions for Algori thmic Analysis at 2011-05-11 12:30:00 DESCRIPTION;LANGUAGE=en-US:Random restrictions are commonly used for provi ng complexity lower bounds, eg. lower bounds on constant-depth circuit s (Ajtai, Furst-Saxe-Sipser, Yao, Hastad) and lower bounds on formula size (Subbotovskaya, Impagliazzo-Nisan, Paterson-Zwick, Hastad). I will show how to use results on random restrictions to bound the running time of certain algorithms for Satisfiability which beat brute-force se arch. Specifically I will present an algorithm which runs in time 2^{n - \Omega(n)} on Boolean formulae of linear size and an algorithm whic h runs in time 2^{n - n^{1/d+1}) on unbounded fan-in circuits of poly(n) size and depth $d$.

I will also mention some more rec ent results motivated by our work and make some general remarks on the co nnections between complexity lower bound techniques and algorithmic analy sis. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013110 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110511T123000 DTEND;TZID="Asia/Jerusalem":20110511T143000 DTSTAMP;TZID="Asia/Jerusalem":20110511T123000 FREEBUSY;FBTYPE=BUSY:20110511T123000/20110511T143000 SUMMARY;LANGUAGE=en-US:The msc talk by Edward Vitkin about Functional geno mics-based approach for reconstructing metabolic network models at 2011-05 -11 12:30:00 DESCRIPTION;LANGUAGE=en-US:Reconstruction of genome-scale metabolic networ ks is considered as key step in quantifying the genotype-phenotype relati onship. A major computational challenge involved in the reconstruction pr ocess is the identification of missing reactions in a metabolic network a process commonly referred to as gap-filling. Here, we present a novel gap-filling approach, MetabolIc Reconstruction via functionAl GEnomics ( MIRAGE) that searches for missing reactions required to catalyze metabolic flux under steady-state whose presence is supported by various function al genomic data. The method follows a two-step procedure. First, functio nal genomics data is utilized to estimate the probability of including e ach reaction from a cross-species reactions database in the final recons tructed network. Then, metabolic flux analysis selects a set of high pro bability reactions, whose addition to the network would enable flux acti vation of all known reactions. The performance of MIRAGE, in comparison to previous methods, is demonstrated in the reconstruction of network mode ls for E. coli and the cyanobacteria Synechocystis. Then, it is applied to reconstruct genome-scale metabolic network models for 36 sequenced cy anobacteria, amenable for constraint-based modeling analysis and specifi cally for metabolic engineering. To demonstrate the utility of the recon structed networks, a strain design method was applied to the reconstruct ed models to predict gene knockouts whose implementation is expected to significantly elevate the production rate of an important nutritional prod uct, astaxanthin. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013130 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110511T140000 DTEND;TZID="Asia/Jerusalem":20110511T160000 DTSTAMP;TZID="Asia/Jerusalem":20110511T140000 FREEBUSY;FBTYPE=BUSY:20110511T140000/20110511T160000 SUMMARY;LANGUAGE=en-US:The msc talk by Igor Kviatkovsky about Color Invari ants for Person Re-Identification at 2011-05-11 14:00:00 DESCRIPTION;LANGUAGE=en-US:We revisit the problem of specific object recog nition using color distributions. In some applications - such as specifi c person identification - it is highly likely that the color distributio ns will be multimodal and hence contain a special structure. We refer to this structure as an intra-distribution structure. We show that th is intra-distribution structure may be used as a signature for recognizi ng objects, that is both discriminative and invariant to a wide range of imaging conditions. Our signature uses shape context descriptors to de scribe this intra-distribution structure. We prove that the signature i s invariant under illumination changes assuming the widely used diagona l model. Our experiments validate the power of our approach. We show good recognition performance using color information only, on several publicly available datasets for person re-identification both indoors and outdoors. We show that our approach combined with existing objec t identification approaches, e.g. the Covariance Descriptor, achieves r esults comparable to the current state of the art on the VIPeR dataset - the most challenging evaluation dataset for person re-identification. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013120 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110512T143000 DTEND;TZID="Asia/Jerusalem":20110512T163000 DTSTAMP;TZID="Asia/Jerusalem":20110512T143000 FREEBUSY;FBTYPE=BUSY:20110512T143000/20110512T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Prof. Peter Pepper SPECIAL GUEST LECTURE about A Unified Formal Approach To Garbage Collection at 2011-05- 12 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013030 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110517T113000 DTEND;TZID="Asia/Jerusalem":20110517T133000 DTSTAMP;TZID="Asia/Jerusalem":20110517T113000 FREEBUSY;FBTYPE=BUSY:20110517T113000/20110517T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Roee Litman (Tel Aviv Univer sity) about Pixel Club: Diffusion-geometric Maximally Stable Component De tection in Deformable Shapes at 2011-05-17 11:30:00 DESCRIPTION;LANGUAGE=en-US:Maximally stable component detection is a very popular method for feature analysis in images, mainly due to its low compu tation cost and high repeatability. With the recent advance of feature-b ased methods in geometric shape analysis, there is significant interest in finding analogous approaches in the 3D world.

In this paper, we formulate a diffusion-geometric framework for stable component detecti on in non-rigid 3D shapes, which can be used for geometric feature detecti on and description. A quantitative evaluation of our method on the SHREC '10 feature detection benchmark shows its potential as a source of high-qu ality features. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013170 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110517T143000 DTEND;TZID="Asia/Jerusalem":20110517T163000 DTSTAMP;TZID="Asia/Jerusalem":20110517T143000 FREEBUSY;FBTYPE=BUSY:20110517T143000/20110517T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Hadas Shachnai about Maximizing Submodular Set Functions Subject to Multiple Linear Constraints at 2011-05 -17 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013200 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110518T103000 DTEND;TZID="Asia/Jerusalem":20110518T123000 DTSTAMP;TZID="Asia/Jerusalem":20110518T103000 FREEBUSY;FBTYPE=BUSY:20110518T103000/20110518T123000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Guy Lebanon (Georgia Institu te of Technology) about Pixel Club: Unsupervised Supervised Learning: Who Needs Labels Anyway? at 2011-05-18 10:30:00 DESCRIPTION;LANGUAGE=en-US:I will describe two recent results in unsupervi sed supervised learning (performing supervised learning tasks without labe ls). The first result concerns evaluating the accuracy of classifiers a nd regression models without labels. The second concerns training margi n based classifiers such as SVM or logistic regression for high dimensiona l data without labels. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013230 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110518T113000 DTEND;TZID="Asia/Jerusalem":20110518T133000 DTSTAMP;TZID="Asia/Jerusalem":20110518T113000 FREEBUSY;FBTYPE=BUSY:20110518T113000/20110518T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Sharon Goldberg (Computer Scienc e Department at Boston University) about ceClub: How Secure are Secu re Internet Routing Protocols? at 2011-05-18 11:30:00 DESCRIPTION;LANGUAGE=en-US:A decade of research has been devoted to addres sing vulnerabilities in global Internet routing system. The result is a pl ethora of security proposals, each providing a different set of security g uarantees. To inform decisions about which proposal should be deployed in the Internet, we present the first side-by-side quantitative comparison of the major security variants. We evaluate security variants on the basis o f their ability to prevent one of the most fundamental forms of attack, wh ere attacker manipulates routing messages in order to attract traffic to a node it controls (so that it can tamper, drop, or eavesdrop on traffic). We combine a graph-algorithmic analysis with simulations on real network d ata to show that prior analysis has underestimated the severity of attacks , even when the strongest known secure routing protocol is fully deployed in the network. We find that simple access control mechanisms can be as ef fective as strong cryptographic approaches, and it is really the combinati on of these two competing approaches that leads to a significant improveme nt in security. Time permitting, we will also discuss some of the economic and engineering issues that must be addressed before any of these proposa ls can realistically be deployed in the Internet.

Based on j oint work with Michael Schapira, Pete Hummon, Jennifer Rexford and Phillip a Gill. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 861 UID:111ec4410400013220 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110518T123000 DTEND;TZID="Asia/Jerusalem":20110518T143000 DTSTAMP;TZID="Asia/Jerusalem":20110518T123000 FREEBUSY;FBTYPE=BUSY:20110518T123000/20110518T143000 SUMMARY;LANGUAGE=en-US:The msc talk by Avihai Mejer about Confidence Estim ation in Structured Prediction at 2011-05-18 12:30:00 DESCRIPTION;LANGUAGE=en-US:Structured classification tasks such as sequenc e labeling and dependency parsing have seen much interest by the Natura l Language Processing and the machine learning communities. Several onl ine learning algorithms were adapted for structured tasks such as Perce ptron, Passive-Aggressive and the recently introduced Confidence-Weight ed learning . These online algorithms are easy to implement, fast to tr ain and yield state-of-the-art performance. However, unlike probabilist ic models like Hidden Markov Model and Conditional random fields, these methods generate models that output merely a prediction with no additi onal information regarding confidence in the correctness of the output. In this work we fill the gap proposing few alternatives to compute the co nfidence in the output of non-probabilistic algorithms. We show how to compute confidence estimates in the prediction such that the confidence reflects the probability that the word is labeled correctly. We then s how how to use our methods to detect mislabeled words, trade recall for precision and active learning. We evaluate our methods on four noun-ph rase chunking and named entity recognition sequence labeling tasks and on dependency parsing for 14 languages. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013190 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110518T123000 DTEND;TZID="Asia/Jerusalem":20110518T143000 DTSTAMP;TZID="Asia/Jerusalem":20110518T123000 FREEBUSY;FBTYPE=BUSY:20110518T123000/20110518T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Pushkar Joglekar (Instit ute of Mathematical Sciences, Chennai, India) about Theory Seminar: Had amard Product of Polynomials and the Identity Testing Problem at 2011-05-1 8 12:30:00 DESCRIPTION;LANGUAGE=en-US: Motivated by the Hadamard product of matrices we define the Hadamard product of noncommutative multivariate polynomials and study its arithmetic circuit and branching program complexity. We also give applications and connections to polynomial identity testing.

One of our main results is a tight characterization of polynomi al identity testing for noncommutative algebraic branching programs over t he field of rationals(we show the problem is complete for logspace countin g class C=L). We also study the complexity of similar identity problem for finite fields, in the case of finite fields we have slightly weaker resul ts.

Next we consider Hadamard product of commutative multivariate polynomials and explore it's expressive power.
ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013210 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110522T113000 DTEND;TZID="Asia/Jerusalem":20110522T133000 DTSTAMP;TZID="Asia/Jerusalem":20110522T113000 FREEBUSY;FBTYPE=BUSY:20110522T113000/20110522T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Dan Feldman (Callifornia Ins titute of Technology) about Pixel Club: Clustering and Approximating High -Dimensional Streaming Data using Coresets at 2011-05-22 11:30:00 DESCRIPTION;LANGUAGE=en-US:Data analysis of massive data sets is common to day for web-search (e.g. Google), social networking (e.g. Facebook), finan cial applications, supermarkets, bioinformatics and many other fields.

A coreset (or, core-set) for a given problem is a "compressed" representation of its input, in the sense that a solution for the pro blem with the (small) coreset as input would yield an approximate solution to the problem with the original (large) input. Using traditional techn iques, a coreset usually implies provable linear time algorithms for the corresponding optimization problem, which can be computed in parallel, via one pass over the data, and using only logarithmic space (i.e, in th e streaming model).

In this talk I will present a unified fra mework that yields the efficient construction of succinct coresets for s everal problems. Representing the data by a set F of positive functions over a ground set X, our framework forges a link between the combinatori al complexity of the function family F at hand (measured in terms of c lassical VC dimension) and the paradigm of coresets. Our coresets are ob tained via randomized sampling according to a delicately designed sampli ng distribution. Examples in which our framework has been successful (a nd significantly improves over previous works) include the k-median, the k-line median, projective clustering, linear regression, low rank appro ximation, and subspace approximation problems. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013270 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110524T113000 DTEND;TZID="Asia/Jerusalem":20110524T133000 DTSTAMP;TZID="Asia/Jerusalem":20110524T113000 FREEBUSY;FBTYPE=BUSY:20110524T113000/20110524T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Artiom Kovnatsky (Applied Ma thematics, Technion) about Pixel Club: Diffusion Framework for Geometric a nd Photometric Data Fusion in Non-rigid Shape Analysis at 2011-05-24 11:30 :00 DESCRIPTION;LANGUAGE=en-US:In this work, we explore the use of the diffusi on geometry framework for the fusion of geometric and photometric informa tion in local and global shape descriptors. Our construction is based on the def- inition of a di ffusion process on the shape manifold embedded i nto a high-dimensional space where the embedding coordinates represent th e photometric information. Experimental results show that such data fu- sion is useful in coping with di fferent challenges of shape analysis wher e pure geometric and pure photometric methods fail ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013290 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110525T113000 DTEND;TZID="Asia/Jerusalem":20110525T133000 DTSTAMP;TZID="Asia/Jerusalem":20110525T113000 FREEBUSY;FBTYPE=BUSY:20110525T113000/20110525T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Gabriel Kliot (MS Research) abou t ceClub: Orleans: A Programming Model for the Cloud at 2011-05-25 11:30:0 0 DESCRIPTION;LANGUAGE=en-US:What if you could build the next Facebook or Tw itter with just a few hundred lines of code ? and scale it to hundreds of millions of users across thousands of servers, right out of the box? Orlea ns is a cloud programming model and runtime from Microsoft Research, which can make this the new "norm".

Orleans is a software framewo rk for building 'client + cloud' applications. It encourages the use of si mple concurrency patterns that are easy to understand and implement correc tly, with an actor-like programming model for distributed applications. It uses declarative specification of persistence, replication, and consisten cy properties, as well as lightweight transactions to support the developm ent of reliable and scalable 'client + cloud' applications.

Bio: Gabriel Kliot a Research Software Development Engineer in the eXtreme Computing Group, Microsoft Research. He's been working on Orleans since i ts early days. Gabriel also leads project Horton: a distributed database f or managing and querying large distributed graphs. Gabriel obtained his Ph D in Computer Science from the Technion in 2009, where his thesis focused on Probabilistic Middleware Services for Wireless Mobile Ad-Hoc Networks, as well as other various aspects of Distributed Systems. In his spare time Gabriel is a competitive long-distance runner. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 3 UID:111ec4410400013250 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110525T123000 DTEND;TZID="Asia/Jerusalem":20110525T143000 DTSTAMP;TZID="Asia/Jerusalem":20110525T123000 FREEBUSY;FBTYPE=BUSY:20110525T123000/20110525T143000 SUMMARY;LANGUAGE=en-US:The msc talk by Michael Kuperstein about Preserving Correctness Under Relaxed Memory Models at 2011-05-25 12:30:00 DESCRIPTION;LANGUAGE=en-US:We present an approach for automatic verificati on of concurrent programs running under relaxed memory models. Verificati on under relaxed memory models is a hard problem. Given a finite state pr ogram and a safety specification, verifying that the program satisfies th e specification under a sufficiently relaxed memory model is undecidable. For somewhat stronger memory models, the problem is decidable but has no n-primitive recursive complexity. We use abstract interpretation to provi de a verification procedure for programs running under relaxed memory mod els. Our main contributions are: -- A family of partial-coherence abstr actions, which partially preserve information required for memory coheren ce and consistency, while allowing effective verification. -- A framew ork for automatic repair of programs. Given a program, a specification an d a description of the memory model, our framework computes a set of cons traints that guarantee the correctness of the program under the memory mo del. The framework then realizes those constraints syntactically as "memo ry fences", hardware instructions that ensure the constraints are never v iolated. We implemented our approach in a tool called BLENDER and used it to infer correct and efficient placements of memory fences for several nontrivial algorithms, including practical concurrent data structures an d mutual exclusion primitives. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013240 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110525T123000 DTEND;TZID="Asia/Jerusalem":20110525T143000 DTSTAMP;TZID="Asia/Jerusalem":20110525T123000 FREEBUSY;FBTYPE=BUSY:20110525T123000/20110525T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Eden Chlamtac (Tel-Aviv University) about Theory Seminar: Linear Index Coding via Semidefinite Pro gramming at 2011-05-25 12:30:00 DESCRIPTION;LANGUAGE=en-US:In the index coding problem, introduced by Birk and Kol (INFOCOM, 1998), the goal is to transmit n bits to n receivers (o ne bit to each), where the receivers reside at the nodes of a graph G an d have prior access to the bits corresponding to their neighbors in the gr aph (side information). The objective is to find a code word of minimum le ngth which will allow each receiver to learn their own bit given access to the code word and their side information. When the encoding is linear (th is is known as linear index coding), the minimum possible code word length corresponds to a graph parameter known as the minrank of G.

In this talk, we will describe an algorithm which approximates the minra nk of a graph in the following sense: when the minrank of the graph is a constant k, the algorithm finds a linear index code of length O(n^(f(k) )). For example, for k=3 we have f(3) ~ 0.2574. This algorithm exploits a connection between minrank and a semidefinite programming (SDP) relaxati on for graph coloring introduced by Karger,Motwani and Sudan.

A result which arises from our analysis, and which may be of independe nt interest, gives an exact expression for the maximum possible value of the Lovasz theta-function of a graph, as a function of its minrank. Thi s compares two classical upper bounds on the Shannon capacity of a graph . ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013280 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110525T143000 DTEND;TZID="Asia/Jerusalem":20110525T163000 DTSTAMP;TZID="Asia/Jerusalem":20110525T143000 FREEBUSY;FBTYPE=BUSY:20110525T143000/20110525T163000 SUMMARY;LANGUAGE=en-US:The phd talk by Zohar Karnin about Explicit Dimensi on Reduction and its Applications at 2011-05-25 14:30:00 DESCRIPTION;LANGUAGE=en-US:We construct a small set of explicit linear tra nsformations mapping R^n to R^t, where t=O(log(\gamma^{-1}) \epsilon^{-2 }), such that the L_2 norm of any vector in R^n is distorted by at most 1 \pm \epsilon in at least a fraction of 1-\gamma of the transformations in the set. Albeit the tradeoff between the size of the set and the suc cess probability is sub-optimal compated with probablistic arguments, we nevertheless are able to apply our construction to a number of problems. In particular, we use it to construct an \epsilon-sample (or pseudo-rand om-generator) for linear threshold functions on S^{n-1}, for \epsilon=o( 1). We also use it to construct an \epsilon-sample for digons in S^{n-1} , for \epsilon=o(1). This construction leads to an efficient oblivious d erandomization of the Goemans-Williamson MAX-CUT algorithm and similar a pproximation algorithms (i.e., we construct a small set of hyperplanes, such that for any instance we can choose one of them to generate a good solution). Our technique for constructing an \epsilon-sample for linear threshold functions on the sphere is considerably different than previo us techniques that rely on k-wise independent sample spaces. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012250 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110530T183000 DTEND;TZID="Asia/Jerusalem":20110530T203000 DTSTAMP;TZID="Asia/Jerusalem":20110530T183000 FREEBUSY;FBTYPE=BUSY:20110530T183000/20110530T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Tomer Ashur about H aifux Club: How to Spread Knowledge Throughout the World While Wearing Onl y Your Slippers (or Wikmedia, Wikipedia and free content projects) at 201 1-05-30 18:30:00 DESCRIPTION;LANGUAGE=en-US:Since its first emergence in 2001, Wikipedia ha d grown drastically to become the fifth most viewed website over the Inter net in 2012. With over 12,000,000 articles in more than 250 languages this is now the largest source of information ever existed. The Wikimedia foun dation has been founded in 2003 as a non-profit organization to support Wi kipedia as well as other online and offline free content projects.

In this talk I will present Wikipedia and other projects of the Wik imedia foundation and how they work. I will show how one can become a Wiki pedia editor and why would she want to do so. We will talk about readers, editors, Neutrality, reliability, dispute resolution, style and free conte nt. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400013160 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110531T143000 DTEND;TZID="Asia/Jerusalem":20110531T163000 DTSTAMP;TZID="Asia/Jerusalem":20110531T143000 FREEBUSY;FBTYPE=BUSY:20110531T143000/20110531T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Gershon Elber about Escher For R eal: on the synergy between science and art at 2011-05-31 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013260 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110601T123000 DTEND;TZID="Asia/Jerusalem":20110601T143000 DTSTAMP;TZID="Asia/Jerusalem":20110601T123000 FREEBUSY;FBTYPE=BUSY:20110601T123000/20110601T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Michael Dinitz (Weizmann Institute) about Theory Seminar: Directed Spanners via Flow-Based Linear Programs at 2011-06-01 12:30:00 DESCRIPTION;LANGUAGE=en-US:A k-spanner of a given graph is a subgraph that preserves all distances within factor k. This notion is useful in sever al contexts, from distributed computing to property testing. By examinin g spanners through flow-based linear programming relaxations, we design an O(n^{2/3})-approximation algorithm for the directed k-spanner problem that works for all k. This is the first sublinear approximation for a rbitrary edge-lengths. We also design a different rounding scheme with a better approximation ratio for the special case of k=3 and unit edge le ngths. Our algorithms easily extend to the fault-tolerant setting, which has recently attracted attention but not from an approximation viewpoin t. A virtue of all of our algorithms is that they are relatively simple.

Joint work with Robert Krauthgamer ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013320 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110601T133000 DTEND;TZID="Asia/Jerusalem":20110601T153000 DTSTAMP;TZID="Asia/Jerusalem":20110601T133000 FREEBUSY;FBTYPE=BUSY:20110601T133000/20110601T153000 SUMMARY;LANGUAGE=en-US:The phd talk by Hanna Mazzawi about Reconstructing Graphs Using Edge Counting Queries at 2011-06-01 13:30:00 DESCRIPTION;LANGUAGE=en-US:In this thesis we study three well known combi natorial search problems in various settings: The coin weighing problem, the problem of reconstructing graphs using additive queries and the probl em of reconstructing hypergraphs using additive queries. All of the th ree combinatorial search problems share a common structure. In each probl em we have a set of objects called a \emph{universe} or an \emph{instance space}. From the instance space a unique object is selected, we call it the \emph{hidden object}. To solve a combinatorial search problem an algo rithm must uniquely identify the hidden object by asking minimal number o f queries of a given type. We distinguish between two types of algorit hms to solve any combinatorial search problem. Adaptive algorithms are al gorithms that take into account outcomes of previous queries while non-a daptive algorithms make all queries in advance, before any answer is know n. In the coin weighing problem, we have a hidden $n$-vector $v$ with $m$ non-zero real number entries. We are allowed to ask queries of the fo rm $$ Q_v(x) = x^T v, $$ where $x\in\{0,1\}^n$. We show the existence of a non-adaptive algorithm for reconstructing the hidden vector with que ry complexity that matches the information theoretic lower bound. We also give a polynomial time adaptive algorithm for the same problem. This al gorithm is optimal for smaller $m$'s and almost optimal for all range of $m$. Moreover, we introduce few techniques that make our algorithms resis tant to noise. That is, using these techniques our algorithms reconstruct correctly the hidden vector even if some of the answers to the queries a re incorrect. In the problem of reconstructing a hidden graph from quer ies, we have a hidden graph $G=(V,E,w)$, where $E\subseteq V\times V$ and $w\in \mathbb{R}^E$. The set of vertices $V$ is known, and the set of e dges $E$ and their weights are unknown. We are allowed to ask queries of the form $$ Q_G(S) = \sum_{e\in E\cap (S\times S)} w(e), $$ for $S\sub seteq V$. That is, the query returns the sum of weights of the edges in t he subgraph induced by $S$. For the problem of reconstructing a hidden we ighted graph, we show the existence of a non-adaptive algorithm with quer y complexity that matches the information theoretic lower bound. We also give the first optimal polynomial time adaptive algorithm for reconstruct ing unweighted graphs. We then extend this result by giving an almost opt imal polynomial time adaptive algorithm for reconstructing weighted grap hs. Finally, we study the problem of reconstructing hidden hypergraphs. We have a hidden hypergraph $H=(V', E', w')$, where $E'\subset 2^{V'}$ and $w'\in \mathbb{R}^{E'}$. As in the previous problem, the set of verti ces is the only set known, and one must reconstruct the hidden hypergraph using queries of the form $$ Q_G(S) = \sum_{e\in E\cap 2^S} w(e), $$ for $S\subseteq V$. That is, the query returns the sum of weights in the subgraph induced by $S$. For this problem, we show the existence of a non -adaptive algorithm for reconstructing an unweighted hypergraph with cons tant rank with query complexity that matches the information theoretic lo wer bound. Here the rank of a hypergraph is the size of its largest edge. We also prove the existence of a non-adaptive algorithm for constant ran k weighted graphs with query complexity that matches the information theo retic lower bound. To achieve the above results, we use numerous techni ques from theoretical computer science such as, the probabilistic method, fourier coefficients analysis of functions, the divide and conquer appr oach, the guess and check approach, methods from coding theory and more. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012000 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110601T150000 DTEND;TZID="Asia/Jerusalem":20110601T170000 DTSTAMP;TZID="Asia/Jerusalem":20110601T150000 FREEBUSY;FBTYPE=BUSY:20110601T150000/20110601T170000 SUMMARY;LANGUAGE=en-US:The msc talk by Jonathan Yaniv about Truthful Mecha nisms for Value-Based Scheduling in Cloud Computing at 2011-06-01 15:00:00 DESCRIPTION;LANGUAGE=en-US:Cloud computing provides an attractive computat ion platform, in which computing resources (e.g., virtual machines, storag e capacity) are rented to end-users under a utility pricing model. The mos t common pricing offering is a pay-as-you-go scheme, in which users are ch arged a fixed price per unit resource per hour. While this scheme is simpl e to implement, it does not support value-based scheduling, where users as sign a value to their job and the cloud's goal is to schedule jobs to maxi mize aggregate value under job demand and capacity constraints. In this work, we introduce a novel pricing and allocation approach for batch jo bs (e.g., image processing, financial analytics, scientific simulations) o n cloud systems. In our economic model, users submit jobs with a value fun ction that specifies willingness to pay as a function of job due dates. Th e cloud provider in response allocates a subset of these jobs, taking into advantage the flexibility of allocating resources to jobs in the cloud en vironment. Focusing on social-welfare as the system objective (especially relevant for private or in-house clouds), we construct a resource allocati on algorithm which obtains a small approximation factor that approaches 2 as the number of servers increases, assuming that user valuations are kn own. Based on this algorithm, we then design an efficient truthful-in-expe ctation mechanism, which significantly improves the running complexity of black-box reduction mechanisms that can be applied to the problem, thereby facilitating its implementation in real systems. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013300 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110602T143000 DTEND;TZID="Asia/Jerusalem":20110602T163000 DTSTAMP;TZID="Asia/Jerusalem":20110602T143000 FREEBUSY;FBTYPE=BUSY:20110602T143000/20110602T163000 SUMMARY;LANGUAGE=en-US:The phd talk by Michael Viderman about Towards lowe r bounds on locally testable codes at 2011-06-02 14:30:00 DESCRIPTION;LANGUAGE=en-US:A Probabilistically Checkable Proof (PCP) is a proof that allows checking the validity of a statement by reading only a constant number of symbols of the proof. The PCP theorem (AS98, ALMSS98 ) asserts the existence of PCPs of polynomial length for any claim that can be stated as membership in an NP set. Surprisingly, all know constructions of PCPs use Locally Testable Codes (LTCs) as their combin atorial core. An LTC is an error-correcting code that has a randomized t ester which reads a constant number of symbols from an input word and de cides whether this word is in the code. It should accept all codewords w ith probability one and reject all words that are far from the code with noticeable probability. The main open question in the area of LTCs is w hether there exists a family of asymptotically good LTCs. Proving the no n-existence for such a family would imply that the current approach for PCP construction can not achieve linear length. In this talk I will describe a new approach to refute the existence of asymptotically good LTCs and report on progress on this problem. Time permitting, I will des cribe a result saying that, allowing sublinear query complexity, one can easily construct a family of LTCs with rate arbitrarily close to 1. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012230 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110613T183000 DTEND;TZID="Asia/Jerusalem":20110613T203000 DTSTAMP;TZID="Asia/Jerusalem":20110613T183000 FREEBUSY;FBTYPE=BUSY:20110613T183000/20110613T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Eli Billauer about Haifux Club: The Anatomy of a PCI/PCI Express Kernel Driver at 2011-06-13 18:30:00 DESCRIPTION;LANGUAGE=en-US:General Purpose Input/Output (GPIO), Serial Per ipheral Interface (SPI), and Inter-Integrated Circuit (I2C), are common me thods for digital communication between electronic components. The Linux k ernel, being a popular choice for embedded solutions, provides a general a bstraction layer for each of those communication methods. Modern Linux ker nels also include drivers for many hardware modules implementing GPIO, SPI , or I2C. The abstraction layers provide a generic way to communicate with electronic devices, which is independent from the details of specific har dware implementation. Each abstraction layer provides API for kernel code, as well as userspace users.

In this talk I'll present the ba sics of each communication method, and its generic userspace interface tha t Linux provides. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400013330 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110614T143000 DTEND;TZID="Asia/Jerusalem":20110614T163000 DTSTAMP;TZID="Asia/Jerusalem":20110614T143000 FREEBUSY;FBTYPE=BUSY:20110614T143000/20110614T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Stefan Wolf about The Price for Perfect Secrecy at 2011-06-14 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013340 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110615T113000 DTEND;TZID="Asia/Jerusalem":20110615T133000 DTSTAMP;TZID="Asia/Jerusalem":20110615T113000 FREEBUSY;FBTYPE=BUSY:20110615T113000/20110615T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Liran Katzir (Yahoo! Israel Labs ) about ceClub: Estimating Sizes of Social Networks via Biased Sampling at 2011-06-15 11:30:00 DESCRIPTION;LANGUAGE=en-US:Online social networks have become very popular in recent years and their number of users is already measured in many hun dreds of millions. For various commercial and sociological purposes, an in dependent estimate of their sizes is important. In this work, algorithms f or estimating the number of users in such networks are considered. The pro posed schemes are also applicable for estimating the sizes of networks' su b-populations.

The suggested algorithms interact with the soc ial networks via their public APIs only, and rely on no other external inf ormation. Due to obvious traffic and privacy concerns, the number of such interactions is severely limited. We therefore focus on minimizing the num ber of API interactions needed for producing good size estimates. We adopt the abstraction of social networks as undirected graphs and use random no de sampling. By counting the number of collisions or non-unique nodes in t he sample, we produce a size estimate. Then, we show analytically that the estimate error vanishes with high probability for smaller number of sampl es than those required by prior-art algorithms. Moreover, although our alg orithms are provably correct for any graph, they excel when applied to soc ial network-like graphs. The proposed algorithms were evaluated on synthet ic as well real social networks such as Facebook, IMDB, and DBLP. Our expe riments corroborated the theoretical results, and demonstrated the effecti veness of the algorithms. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Meyer 861 UID:111ec4410400013430 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110615T123000 DTEND;TZID="Asia/Jerusalem":20110615T143000 DTSTAMP;TZID="Asia/Jerusalem":20110615T123000 FREEBUSY;FBTYPE=BUSY:20110615T123000/20110615T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Sanjam Garg, (UCLA) abou t Theory Seminar: Leakage-Resilient Zero-Knowledge at 2011-06-15 12:30:00 DESCRIPTION;LANGUAGE=en-US:We initiate a study of zero knowledge proof sys tems in the presence of side-channel attacks. Specifically, we consider a setting where a cheating verifier is allowed to obtain arbitrary bounded l eakage on the entire state (including the witness and the random coins) of the prover during the entire protocol execution. We formalize a meaningfu l definition of leakage-resilient zero knowledge (LR-ZK) proof system, tha t intuitively guarantees that the protocol does not yield anything beyond the validity of the statement and the leakage obtained by the verifier.

We give a construction of LR-ZK interactive proof system bas ed on general assumptions. To the best of our knowledge, this is the first instance of a cryptographic interactive protocol where the adversary is a llowed to perform leakage attacks on the entire state of honest party duri ng the protocol execution (in contrast, prior work only considered leakage prior to the protocol execution, or very limited leakage during during th e protocol execution). Next, we give an LR-NIZK argument system based on s tandard assumptions.

Finally, we demonstrate the usefulnes s of our notions by giving two concrete applications: -- We show how to do UC secure computation in the "leaky token model" (i.e., where an advers ary in possession of a token can obtain arbitrary bounded leakage on the e ntire state of the token) based on standard assumptions. -- Next, we giv e a new construction of fully leakage-resilient signatures in the bounded leakage model (as well as the continual leakage model) based on standard a ssumptions. In contrast to the recent constructions of fully leakage resil ient signatures, our scheme is also secure in the "noisy leakage" model.

Joint work with Abhishek Jain and Amit Sahai. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013410 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110615T133000 DTEND;TZID="Asia/Jerusalem":20110615T153000 DTSTAMP;TZID="Asia/Jerusalem":20110615T133000 FREEBUSY;FBTYPE=BUSY:20110615T133000/20110615T153000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Yonathan Savir (Ph ysics of Complex Systems, Weizmann Institute) about Bioinformatics Foru m: Molecular Recognition in Presence of Competition and Noise at 2011-06-1 5 13:30:00 DESCRIPTION;LANGUAGE=en-US:Molecular recognition plays a key role in proce ssing information in biological systems which rely on the ability of bio -molecules to specifically recognize each other. However, the crowded bi ological environment contains a vast variety of molecules that are often structurally similar and may compete with the "right" target. Thus, the recognition process is prone to false binding, which introduces errors and may impair the proper information flow and must be taken into accoun t in the design and the evolution of molecular information channels. To evaluate the optimal design of a molecular recognizer, we formulated mol ecular recognition as a transmission of information via a noisy channel. Using this framework, I will discuss few essential molecular recognitio n systems: 1) Decoding of tRNA by the ribosome during translation 2) Hom ologous recombination, the process in which two identical or similar DNA molecules exchange genetic material and 3) CO2 fixation during photosyn thesis which is mediated by the enzyme Rubisco, probably the most abunda nt protein on Earth. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400013310 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110615T140000 DTEND;TZID="Asia/Jerusalem":20110615T160000 DTSTAMP;TZID="Asia/Jerusalem":20110615T140000 FREEBUSY;FBTYPE=BUSY:20110615T140000/20110615T160000 SUMMARY;LANGUAGE=en-US:The msc talk by Yuval Shimron about Smaller Footpri nt for Java Collections at 2011-06-15 14:00:00 DESCRIPTION;LANGUAGE=en-US:In dealing with the container bloat problem, we identify five memory compaction techniques, which can be used to reduce the footprint of the small objects that make these containers. Using thes e techniques, we describe two alternative methods for more efficient encod ing of the JRE's ubiquitous HashMap data structure, and present a mathema tical model in which the footprint of this can be analyzed. First of this is our fused-hashing encoding method, which reduces memory overhead b y 20%-40% on a 32-bit environment and 45%-65% on a 64-bit environment. T his encoding guarantees these figures as lower bound regardless of the d istribution of keys in hash buckets. Second, our more opportunistic squa shed-hashing, achieves expected savings of 25%-70% on a 32-bit environme nt and 30%-75% on a 64-bit environments, but these savings can degrade and are not guaranteed against bad (and unlikely) distribution of keys to b uckets. Timing results indicate that squashed hashing generally improve s the SPECjbb2005 benchmark, but some of the operations, particularly re movals and iterations are slow. We still find squashed hashing is faster then the baseline implementation for large tables indexed by Integers. We note that with the use of our compaction techniques, there is merit in an implementation of HashSet which is distinct from that of HashMap. For TreeMap we show two encodings which reduces the overhead of tree node s by ~43%, ~46% on a 32-bit environment and ~55%, ~73% on a 64-bit envir onment. These compactions also give a reason to separating the implement ation of TreeSet from that of TreeMap. A separete implementation is expec ted to increase the footprint reduction to ~59%, ~54% on a 32-bit enviro nment and ~61%, ~75% on a 64-bit environment. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013370 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110615T160000 DTEND;TZID="Asia/Jerusalem":20110615T180000 DTSTAMP;TZID="Asia/Jerusalem":20110615T160000 FREEBUSY;FBTYPE=BUSY:20110615T160000/20110615T180000 SUMMARY;LANGUAGE=en-US:The msc talk by Tom Kolan about Coding Techniques f or Burst Errors at 2011-06-15 16:00:00 DESCRIPTION;LANGUAGE=en-US:List decoding of error-correcting codes is a ge neralization of unique decoding: while unique decoding relates to the cas e where the decoder outputs only one word (the correct codeword), list de coding allows to output a list of codewords, as long as the correct codew ord is included in the list. Codes for burst error correction have been s tudied mainly for the purpose of unique decoding. Understanding list deco ding of burst errors is desirable for reliable delivery of data over vari ous unreliable communication channels. In this work, we derive new bou nds for list decodable burst error correcting codes, as well as codes ach ieving these bounds. The bounds are genralizations of the Reiger bound fo r burst error correcting codes, which states that a code is able to corre ct all \tau-bursts only if its redundancy is at least 2\tau. We find boun ds for group codes (e.g. linear codes) and also for general block codes. We present explicit constructions of codes achieving the bounds. Furtherm ore, we present explicit constructions of codes achieving the previously known bounds having higher code lengths than the known codes. Decoding of burst errors is a fairly straight forward and low complexity algorithm, so the codes can be used in actual communication systems. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:in Taub 601 UID:111ec4410400013380 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110616T103000 DTEND;TZID="Asia/Jerusalem":20110616T123000 DTSTAMP;TZID="Asia/Jerusalem":20110616T103000 FREEBUSY;FBTYPE=BUSY:20110616T103000/20110616T123000 SUMMARY;LANGUAGE=en-US:The cggc talk by Liang Liu (Zhejiang University, Ha ngzhou, China) about CGGC Seminar: Geometry-driven Image Manipulation at 2011-06-16 10:30:00 DESCRIPTION;LANGUAGE=en-US:The talk will include two parts. In the first p art, I will present an overview of my research in geometry processing, i ncluding mesh parameterization, surface reconstruction, shape analysis a nd segmentation. In the second part, I will describe my recent work on geo metry-driven image manipulation, including mesh-warping based image reta rgeting, photo composition optimization and parametric human reshaping. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013350 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110616T103000 DTEND;TZID="Asia/Jerusalem":20110616T123000 DTSTAMP;TZID="Asia/Jerusalem":20110616T103000 FREEBUSY;FBTYPE=BUSY:20110616T103000/20110616T123000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Ligang Liu ( Zhejiang Univer sity, Hangzhou, China) about Pixel Club: Geometry-driven Image Manipulatio n at 2011-06-16 10:30:00 DESCRIPTION;LANGUAGE=en-US:The talk will include two parts. In the first p art, I will present an overview of my research in geometry processing, inc luding mesh parameterization, surface reconstruction, shape analysis and segmentation. In the second part, I will describe my recent work on geo metry-driven image manipulation, including mesh-warping based image reta rgeting, photo composition optimization and parametric human reshaping. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013420 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110619T143000 DTEND;TZID="Asia/Jerusalem":20110619T163000 DTSTAMP;TZID="Asia/Jerusalem":20110619T143000 FREEBUSY;FBTYPE=BUSY:20110619T143000/20110619T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Amit Sahai about Efficient and E xplicit Coding for Interactive Communication at 2011-06-19 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013400 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110621T113000 DTEND;TZID="Asia/Jerusalem":20110621T133000 DTSTAMP;TZID="Asia/Jerusalem":20110621T113000 FREEBUSY;FBTYPE=BUSY:20110621T113000/20110621T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Yair Weiss (Hebrew Universit y of Jerusalem) about Pixel Club: Semi-Supervised Learning in Gigantic Ima ge Collection at 2011-06-21 11:30:00 DESCRIPTION;LANGUAGE=en-US:With the advent of the Internet it is now possi ble to collect hundreds of millions of images for computer vision. These images come with varying degrees of label information. "Clean" labels c an be manually obtained on a small fraction, "noisy labels" may be extra cted automatically from surrounding text, while for most images there ar e no labels at all. Semi-supervised learning is a principled framework for combining these different label sources. However, it scales polynom ially with the number of images, making it impractical for use on gigant ic collections with hundreds of millions of images and thousands of clas ses. In this paper we show how to utilize recent results in machine lear ning to obtain highly efficient approximations for semi-supervised learn ing that are linear in the number of images. Specifically, we use the co nvergence of the eigenvectors of the normalized graph Laplacian to eigen functions of weighted Laplace-Beltrami operators. Our algorithm enables us to apply semi-supervised learning to a database of 80 million images gathered from the Internet. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400013450 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110621T143000 DTEND;TZID="Asia/Jerusalem":20110621T163000 DTSTAMP;TZID="Asia/Jerusalem":20110621T143000 FREEBUSY;FBTYPE=BUSY:20110621T143000/20110621T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Paul H. Siegel about Information Rates for Channels with Synchronization Errors at 2011-06-21 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400013390 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110622T113000 DTEND;TZID="Asia/Jerusalem":20110622T133000 DTSTAMP;TZID="Asia/Jerusalem":20110622T113000 FREEBUSY;FBTYPE=BUSY:20110622T113000/20110622T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Ariel Orda (EE Technion) about c eClub: Network Science - A Network of Sciences at 2011-06-22 11:30:00 DESCRIPTION;LANGUAGE=en-US:Network Science is a newly emerging discipline with applications in a variety of domains, such as Communication Networks, Power Grid Networks, Transportation Networks, Social Networks and Biologi cal Networks. Focusing on communication networks, we shall discuss what ne twork science should be and what it should consist of. The talk will also feature some historical anecdotes, tracing back to ancient times. Further details would be a spoiler. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Meyer 861 UID:111ec4410400013500 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110622T133000 DTEND;TZID="Asia/Jerusalem":20110622T153000 DTSTAMP;TZID="Asia/Jerusalem":20110622T133000 FREEBUSY;FBTYPE=BUSY:20110622T133000/20110622T153000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Prahladh Harsha (Tata In stitute of Fundamental Research,Mumbai, India) about Theory Seminar: Al most Settling the Hardness of Noncommutative Determinant at 2011-06-22 13 :30:00 DESCRIPTION;LANGUAGE=en-US:The determinant and the permanent of a matrix, though deceivingly similar in their definitions, behave very differently w ith respect to how efficiently one can compute these quantities. The deter minant of a matrix over a field can be easily computed via Gaussian elimin ation while computing the permanent, as shown by Valiant, is at least as h ard as counting the number of satisfiable assignments to a Boolean formula . Given this, it is natural to ask "over which algebras, is the determin ant easier to compute than the permament? Furthermore, since all algorithm s for determinant crucially use commutivity of the underlying algebra, w e could ask "is commutativity essential for efficient determinant computat ion?"

Extending the recent result of Arvind and Srinivasan [S TOC 2010], we show that computing the determinant of an nxn matrix whose e ntries are themselves 2x2 matrices over a field is as hard as computing th e permanent over the field. On the other hand, surprisingly if one restric ts the elements to be dxd upper triangular matrices, then determinant can be computed in poly(n^d) time. Combining this with the decomposition th eorem of finite algebras, we get the following dichotomy result: if A is a constant dimensional algebra over a finite field of odd characteristic, t hen the commutativity of the quotient algebra A/R(A) determines efficient determinant computation (where R(A) is the radical of A).

Thi s is joint work with Steve Chien, Alistair Sinclair and Srikanth Srinivasa n. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013490 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110622T153000 DTEND;TZID="Asia/Jerusalem":20110622T173000 DTSTAMP;TZID="Asia/Jerusalem":20110622T153000 FREEBUSY;FBTYPE=BUSY:20110622T153000/20110622T173000 SUMMARY;LANGUAGE=en-US:The msc talk by Alon Mishne about PRIME - Programmi ng with Millions of Examples at 2011-06-22 15:30:00 DESCRIPTION;LANGUAGE=en-US:We present the PRIME tool which utilizes static specification mining techniques to extract useful specifications of libra ry APIs from a large number of code fragments that use it, and then uses d ata mining techniques to aggregate the samples into use-cases and sort the m according to popularity and complexity. Programming is becoming mor e and more about using frameworks and libraries, with most of them designe d to support a wide range of usage scenarios. Typically, a programmer only needs partial functionality from a library, and is required to navigate t he extensive library interface (API) to find how to implement the desired functionality. Instead of navigating the complicated library code and do cumentation, programmers often rely on code examples of client programs th at use the library. Such code examples can be obtained from library docume ntation, from other programmers, or via a myriad of search engines and oth er online tools. The availability of services such as Google Code Search exposes the programmer to a vast number of code examples. Making sense of these examples, however, can be an extremely challenging task. Code fragm ents using the API of interest may appear in slightly different contexts a nd are often interleaved with irrelevant code, making it hard for a progra mmer to tease out the relevant details. Furthermore, for a given code samp le there is always the possibility its use of the API is erroneous or sub- optimal. These factors make it hard for a human to benefit from this vast amount of available information. Using a combination of program analysis and machine learning techniques, PRIME mines library specifications from a large collection of client code using it, allowing programmers to write new code using the library, even when they are not familiar with it. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013480 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110626T130000 DTEND;TZID="Asia/Jerusalem":20110626T150000 DTSTAMP;TZID="Asia/Jerusalem":20110626T130000 FREEBUSY;FBTYPE=BUSY:20110626T130000/20110626T150000 SUMMARY;LANGUAGE=en-US:The cggc talk by Myung-Soo Kim (School of Computer Science and Engineering, Seoul National University, Seoul, Korea) about CG GC Seminar: Efficient Algorithms for Freeform Geometric Models at 2011-06- 26 13:00:00 DESCRIPTION;LANGUAGE=en-US:We present a new approach to the development of efficient geometric algorithms for freeform curves and surfaces. Prepr ocessing the given curves and surfaces and representing them in a hiera rchical data structure, we show that a variety of geometric algorithms can be greatly accelerated. We demonstrate the effectiveness of this ap proach by developing real-time algorithms for collision detection, minimum and Hausdorff distance computation, convex hull computation for freefor m models.

This is a joint work with Yong-Joon Kim and Gershon Elber. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013440 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110627T113000 DTEND;TZID="Asia/Jerusalem":20110627T133000 DTSTAMP;TZID="Asia/Jerusalem":20110627T113000 FREEBUSY;FBTYPE=BUSY:20110627T113000/20110627T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Kenneth Weiss (Computer Scie nce, University of Maryland) about Pixel Club: Diamond-based Models for Sc ientificVisualization at 2011-06-27 11:30:00 DESCRIPTION;LANGUAGE=en-US: will describe some recent work towards scaling the process of visual recognition to handle thousands of objects classe s on a limited computational budget. Formally, the system is designed su ch that the computational resources grow sub-linearly (poly-logarithmic) with the number of classes. This also implies that features used for re cognition should be shared by several classes.

Work done jo intly with Shai Shalev-Schwartz and Yonatan Wexler ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400013540 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110627T183000 DTEND;TZID="Asia/Jerusalem":20110627T203000 DTSTAMP;TZID="Asia/Jerusalem":20110627T183000 FREEBUSY;FBTYPE=BUSY:20110627T183000/20110627T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Baruch Siach (TK Op en Systems) about Haifux Club: GPIO, SPI, and I2C Control from Userspace, the True Linux Way at 2011-06-27 18:30:00 DESCRIPTION;LANGUAGE=en-US:General Purpose Input/Output (GPIO), Serial Per ipheral Interface (SPI), and Inter-Integrated Circuit (I2C), are common me thods for digital communication between electronic components. The Linux k ernel, being a popular choice for embedded solutions, provides a general a bstraction layer for each of those communication methods. Modern Linux ker nels also include drivers for many hardware modules implementing GPIO, SPI , or I2C. The abstraction layers provide a generic way to communicate with electronic devices, which is independent from the details of specific har dware implementation. Each abstraction layer provides API for kernel code, as well as userspace users.

In this talk I'll present the ba sics of each communication method, and its generic userspace interface tha t Linux provides. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400013470 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110628T113000 DTEND;TZID="Asia/Jerusalem":20110628T133000 DTSTAMP;TZID="Asia/Jerusalem":20110628T113000 FREEBUSY;FBTYPE=BUSY:20110628T113000/20110628T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Amnon Shashua (Hebrew Univer sity of Jerusalem) about Pixel Club: Scalability of Visual Recognition: F itting Computational Resources for the Task at 2011-06-28 11:30:00 DESCRIPTION;LANGUAGE=en-US:Hierarchical spatial decompositions are a basic modeling tool in a variety of application domains including scientific visualization, finite element analysis and shape modeling and analysis. A popular class of such approaches is based on the regular simplex bisec tion operator, which bisects simplices (e.g. line segments, triangles, t etrahedra) along the midpoint of a predetermined edge. Regular simplex b isection produces adaptive simplicial meshes of high geometric quality, while simplifying the extraction of crack-free, or conforming, approxima tions to the original dataset. Efficient multiresolution representations for such models have been achieved in 2D and 3D by clustering sets of s implices sharing the same bisection edge into structures called diamonds .

In this talk, we formalize the notion of diamonds in arbitrary dimensions in terms of two related simplicial decompositions of hypercu bes leading to a compact pointerless representation for conforming meshe s generated by regular simplex bisection. We then introduce the supercub e as a high-level primitive on such nested meshes based on the atomic un its within the underlying triangulation grid. We discuss the use of supe rcubes to associate information with coherent subsets of the full hierar chy and demonstrate the effectiveness of this representation for modelin g multiresolution terrain and volumetric datasets. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400013530 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110628T123000 DTEND;TZID="Asia/Jerusalem":20110628T143000 DTSTAMP;TZID="Asia/Jerusalem":20110628T123000 FREEBUSY;FBTYPE=BUSY:20110628T123000/20110628T143000 SUMMARY;LANGUAGE=en-US:The phd talk by Natalia Silberstein about Coding Th eory and Projective Spaces at 2011-06-28 12:30:00 DESCRIPTION;LANGUAGE=en-US:The projective space of order n over the finite field F is the set of all the subspaces of the vector space F^n. A code C in the projective space is defined as a subset of the projective space, i .e., the codewords in C are subspaces of F^n. If all the codewords in C ha ve the same dimension, then C is called a constant dimension code. These c odes gained renewed interest due to the work by Koetter and Kschischang (2 008), where they presented an application of such codes to error correctio n in random network coding. In the first part of the talk we will pres ent a method to design error-correcting codes in the projective space. We use a multilevel approach to design our codes. First, we select a constant weight code. Each codeword defines a skeleton of a basis for a subspace i n reduced row echelon form. This skeleton contains a Ferrers diagram on wh ich we design a rank-metric code. Each such rank-metric code is lifted to a constant dimension code. The union of these codes is our final constant dimension code. In particular the codes constructed by Koetter and Kschisc hang are a subset of our codes. In the second part of the talk we will discuss a relationship between constant dimension codes and combinatorial designs. We consider lifted maximum rank distance codes. We first prove t hat each such code forms a transversal design in sets and it also forms a structure akin to transversal design in subspaces. We present bounds on th e size of codes which contain lifted maximum rank distance codes and descr ibe constructions of codes which attain these bounds. Finally, we present LDPC codes based on the lifted maximum rank distance codes and discuss the ir properties. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013510 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110629T123000 DTEND;TZID="Asia/Jerusalem":20110629T143000 DTSTAMP;TZID="Asia/Jerusalem":20110629T123000 FREEBUSY;FBTYPE=BUSY:20110629T123000/20110629T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Shubhangi Saraf (MIT and IAS) about Theory Seminar: High Rate Error Correcting Codes with Sublinea r Time Decoding at 2011-06-29 12:30:00 DESCRIPTION;LANGUAGE=en-US:Locally decodable codes are error-correcting co des that admit efficient decoding algorithms: They give a method to enc ode k bit messages into n bit codewords such that even after a constant fraction of the bits of the codeword get corrupted any bit of the orig inal message can be recovered by only looking at q(k) bits of the corru pted codeword. The tradeoff between the rate of a code (i.e., the ratio k/n) and the locality/efficiency (the function q(k)) of its decoding a lgorithms has been studied extensively in the past decade. However mos t prior work has focused on codes with extremely small q (e.g., constant functions), and the resulting constructions suffer from extremely poor r ate (k/n goes to zero; in fact n grows nearly exponentially in k). Inde ed it was widely believed that such behavior is inherent: Known codes w ith locality k^{1/c} had rate 2^{- O(c)}; and no code with non-trivial locality (q = o(k)) and rate > 1/2 (the practical regime) was known.

In this talk we overcome the rate 1/2 barrier and give a new c lass of codes with very high rates (arbitrarily close to 1) with strong local decoding properties (q(k) = k^{epsilon} for arbitrarily small ep silon), thereby giving new performance tradeoffs between the rate and l ocality of decoding. These codes, which we call multiplicity codes, ar e based on evaluating high degree multivariate polynomials and their de rivatives. Multiplicity codes extend traditional multivariate polynomia l based codes; they inherit the local-decodability of these codes, and at the same time achieve better tradeoffs and flexibility in the rate a nd decodability. These codes give respectable parameters even in concre te settings (and not just in asymptotic behavior), giving hope that loc al decoding techniques may have practical implications.

Base d on joint work with Swastik Kopparty (IAS) and Sergey Yekhanin (MSR, S ilicon Valley) ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 9 UID:111ec4410400013560 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110629T133000 DTEND;TZID="Asia/Jerusalem":20110629T153000 DTSTAMP;TZID="Asia/Jerusalem":20110629T133000 FREEBUSY;FBTYPE=BUSY:20110629T133000/20110629T153000 SUMMARY;LANGUAGE=en-US:The msc talk by Pavel Gurevich about Experimental C lassical Alice Quantum Key Distribution Protocol at 2011-06-29 13:30:00 DESCRIPTION;LANGUAGE=en-US:We present the first experimental realization o f a new semi-classical quantum key distribution (QKD) protocol called cla ssical Alice (with mirror) and its security analysis. The field of qua ntum information and computation is relatively new and rapidly emerging f ield of science, and secure key distribution is one of the most prominent practical applications in this area. Unlike other conventional key distr ibution schemes --- to which we refer as classical --- that rely on (not fully proved) assumptions in mathematics and complexity theory, quantum k ey distribution relies on quantum physics imposed rules to gain its secur ity. Secure key distribution scheme based on quantum carriers (quantum bits) was first proposed by Bennett and Brassard in 1984. This scheme was later proved to be secure, and its security was proved against the most general kind of attacks thus making schemes using this protocol informati on secure. originally, quantum key distribution schemes demanded that both parties have quantum abilities. In 2007, it was shown (at the Techni on) that the requirement for both parties "quantumness" can be relaxed wi thout violating security. The first protocol, QKD with classical Bob, w as followed by a simpler one, QKD with classical Alice which we have impl emented and analyzed. We describe a free space based setup that uses visi ble light to implement a variation of the classical Alice protocol. Our s etup which employs a novel idea (suggested by Matty Katz) of a "sub-qubit " measurement using the rotating mirror approach. This idea enables pr actical implementations of Alice side, and furthermore also renders sever al attacks again real-world implementations of the original classical Ali ce protocol impossible. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013520 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110630T123000 DTEND;TZID="Asia/Jerusalem":20110630T143000 DTSTAMP;TZID="Asia/Jerusalem":20110630T123000 FREEBUSY;FBTYPE=BUSY:20110630T123000/20110630T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Arnab Bhattacharyya (MIT ) about Theory Seminar: Testing Odd-Cycle-Freeness in Boolean Functions at 2011-06-30 12:30:00 DESCRIPTION;LANGUAGE=en-US:Call a function f : {0,1}^n -> {0,1} odd-cycle- free if there are no x_1, ..., x_k in {0,1}^n with k an odd integer such t hat f(x_1) = ... = f(x_k) = 1 and x_1 + ... + x_k = 0. We show that one ca n distinguish odd-cycle-free functions from those eps-far from being odd-c ycle-free by making poly(1/eps) queries to an evaluation oracle. To obtain this result, we use connections between Fourier analysis and spectral gra ph theory to show that one can reduce testing odd-cycle-freeness of Boolea n functions to testing bipartiteness of dense graphs. Our results imply th at one can estimate the smallest Fourier coefficient of a function upto ad ditive error eps by making poly(1/eps) queries.

Our work form s part of a recent sequence of works that shows connections between testab ility of properties of Boolean functions and of graph properties. We also prove that there is a canonical tester for odd-cycle-freeness making poly( 1/eps) queries, meaning that the testing algorithm operates by picking a r andom linear subspace of dimension O(log 1/eps) and then checking if the r estriction of the function to the subspace is odd-cycle-free or not. The t est is analyzed by studying the effect of random subspace restriction on t he Fourier coefficients of a function. Our work implies that testing odd-c ycle-freeness using a canonical tester instead of an arbitrary tester incu rs no more than a polynomial blowup in the query complexity. The question of whether a canonical tester with polynomial blowup exists for all linear -invariant properties remains an open problem.

Joint work wi th Elena Grigorescu, Prasad Raghavendra, and Asaf Shapira. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013570 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110703T130000 DTEND;TZID="Asia/Jerusalem":20110703T150000 DTSTAMP;TZID="Asia/Jerusalem":20110703T130000 FREEBUSY;FBTYPE=BUSY:20110703T130000/20110703T150000 SUMMARY;LANGUAGE=en-US:The cggc talk by Uri Itai (Applied Mathematics, Tec hnion) about CGGC Seminar: An Interpolatory Subdivision Scheme for Positiv e Definite Matrices at 2011-07-03 13:00:00 DESCRIPTION;LANGUAGE=en-US:Symmetric positive definite matrices are widel y used. Diffusion kernels, Curvature, Optimization and many more. Howeve r, Symmetric positive definite matrices are not a group and form an open unbounded manifold. Thus, not clear how to efficiently interpolate such d ata. In this lecture we give such a subdivision interpolation scheme base d on the geodes in the Rieman metric of the Symmetric positive definite matrices. We prove convergences and smoothness and in addition spectral properties as well.

This is a joint work with Nir Sharon and Nira Dyn. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013360 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110706T113000 DTEND;TZID="Asia/Jerusalem":20110706T133000 DTSTAMP;TZID="Asia/Jerusalem":20110706T113000 FREEBUSY;FBTYPE=BUSY:20110706T113000/20110706T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Amit Aides (EE, Technion) ab out Pixel Club: Multiscale Ultrawide Foveated Video Extrapolation at 2011- 07-06 11:30:00 DESCRIPTION;LANGUAGE=en-US:Video extrapolation is the task of extending a video beyond its original field of view. Extrapolating video in a manner that is consistent with the original video and visually pleasing is dif ficult. In this work we aim at very wide video extrapolation which incre ases the complexity of the task. Some video extrapolation methods simpli fy the task by using a rough color extrapolation. A recent approach focu ses on artifact avoidance and run time reduction using foveated video ex trapolation, but fails to preserve the structure of the scene. This talk introduces a multi-scale method which combines a coarse to fine approac h with foveated video extrapolation. Foveated video extrapolation reduce s the effective number of pixels that need to be extrapolated, making the extrapolation less time consuming and less prone to artifacts. The coars e to fine approach better preserves the structure of the scene while pre serving finer details near the domain of the input video. The combined m ethod gains improvement both visually and in processing time.

M.Sc. under the supervision of Prof. Yoav Y. Schechner. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400013580 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110706T113000 DTEND;TZID="Asia/Jerusalem":20110706T133000 DTSTAMP;TZID="Asia/Jerusalem":20110706T113000 FREEBUSY;FBTYPE=BUSY:20110706T113000/20110706T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Panagiota Fatourou (Forth-ICS) a bout ceClub: Highly Efficient Synchronization Techniques at 2011-07-06 11: 30:00 DESCRIPTION;LANGUAGE=en-US:We present highly-efficient synchronization tec hniques and experimentally show that these techniques outperform most stat e of the art lock-based and lock-free synchronization mechanisms. One of t he techniques ensures, in addition, wait-freedom.

We have us ed these techniques to implement common concurrent data structures, like s tacks and queues. Our experiments show that these data structure implement ations have much better performance than state of the art shared stack and queue implementations which ensure only weaker progress properties than o ne of our techniques.

Bio: Panagiota Fatourou is an Assistan t Professor at the Department of Computer Science, University of Crete and an affiliated faculty member of the Institute of Computer Science (ICS) o f the Foundation for Research and Technology-Hellas (FORTH). Prior to join ing the University of Crete and FORTH ICS, she was a full-time faculty mem ber at the Department of Computer Science of the University of Ioannina. T he academic years 2000 and 2001, she was a postdoc at Max-Planck Institut fur Informatik, Saarbrucken, Germany, and at the Computer Science Departme nt of the University of Toronto, Canada. She got a degree in Computer Scie nce from the University of Crete, and a PhD degree in Computer Engineering from the University of Patras. Her research interests focus on the theory of parallel and distributed computing. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 3 UID:111ec4410400013620 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110706T123000 DTEND;TZID="Asia/Jerusalem":20110706T143000 DTSTAMP;TZID="Asia/Jerusalem":20110706T123000 FREEBUSY;FBTYPE=BUSY:20110706T123000/20110706T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Swastik Kopparty (IAS an d Rutgers) about Theory Seminar: On the Complexity of Powering in Finite F ields at 2011-07-06 12:30:00 DESCRIPTION;LANGUAGE=en-US:We study the complexity of powering in GF(2^n) by constant depth arithmetic circuits over GF(2) (also known as AC0(pari ty)). Our study encompasses basic arithmetic operations such as computin g cube-roots and cubic-residuosity of elements of GF(2^n). Our main resu lt is that these operations require exponential size circuits.

We also derive strong average-case versions of these results. For exam ple, we show that no subexponential-size, constant-depth, arithmetic cir cuit over GF(2) can correctly compute the cubic residue symbol for more than 1/3 + o(1) fraction of the elements of GF(2^n). As a corollary, we deduce a character sum bound showing that the cubic residue character ov er GF(2^n) is uncorrelated with all degree-d n-variate GF(2) polynomials (viewed as functions over GF(2^n) in a natural way), provided d << n^{0 .1}.

Our proof revisits the classical Razborov-Smolensky meth od for circuit lower bounds, and executes an analogue of it in the land of univariate polynomials over GF(2^n). ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013600 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110706T130000 DTEND;TZID="Asia/Jerusalem":20110706T150000 DTSTAMP;TZID="Asia/Jerusalem":20110706T130000 FREEBUSY;FBTYPE=BUSY:20110706T130000/20110706T150000 SUMMARY;LANGUAGE=en-US:The msc talk by Elior Malul about Correctness of de rived events at 2011-07-06 13:00:00 DESCRIPTION;LANGUAGE=en-US:Event processing has become a popular infrastru cture for applications which require real-time processing of high-volume d ata streams, such as electronic trading systems, and network and infrast ructure monitoring. Although many modern systems push the limits of tradi tional data processing systems to ever new heights both in terms of perfo rmance and productivity, they do little to ensure the consistency of thei r result. In some cases, event processing applications suffer from incorr ect results, since the order in which events arrive to event processing a gents is unknown. In this paper we propose a solution for one of the majo r aspect of correctness; the ability to define and enforce correctness r equirements on events that are emitted as a result of event processing fun ctions operations, such as filtering and aggregations. Based on observa tions within a significant amount of applications, we define three correc tness schemes that can be used by the system designer when designing an ev ent processing application. An example is the fairness scheme, which guar antees that events are processed in their correct logical order, regardl ess of their preparation path. The paper is accompanied by two scenarios used to demonstrate these correctness schemes. Using these scenarios, we explain, demonstrate, and then formally define each scheme. While the co rrectness schemes are model independent, we further define the extension s needed in the event processing model, in order to support these schemes. Finally, we present two run-time algorithms to enforce the schemes in t he model. The work in this paper opens the gate for development construct abstractions, that have the potential to ensure deterministic results of event processing applications. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013550 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110713T103000 DTEND;TZID="Asia/Jerusalem":20110713T123000 DTSTAMP;TZID="Asia/Jerusalem":20110713T103000 FREEBUSY;FBTYPE=BUSY:20110713T103000/20110713T123000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Ronen Shaltiel (Haifa Un iversity) about Theory Seminar: Dispersers for Affine Sources with Sub-Pol ynomial Entropy at 2011-07-13 10:30:00 DESCRIPTION;LANGUAGE=en-US:We construct an explicit disperser for affine s ources over $\F_2^n$ with entropy $k=2^{\log^{0.9} n}=n^{o(1)}$. This is a polynomial time computable function $D:\F_2^n \to \set{0,1}$ such that fo r every affine space $V$ of $\F_2^n$ that has dimension at least $k$, $D(V )=\set{0,1}$. This improves the best previous construction of Ben-Sasson a nd Kopparty that achieved $k = \Omega(n^{4/5})$ and is the first pseudoran dom object for affine sources with entropy less than $\sqrt{n}$.
Our technique follows a high level approach that was introduced by Ba rak, Kindler, Shaltiel, Sudakov and Wigderson, and further developed by Ba rak, Rao, Shaltiel and Wigderson, in the context of dispersers for two in dependent general sources. We implement a "challenge-response game" for af fine sources. This game allows the disperser to locate high entropy blocks in the source (even though the disperser only receives a single sample fr om the source). In order to implement the game we rely (amongst other thin gs) on ideas of Rao in the context of extractors for low-weight affine sou rces. We also use a recursive win-win analysis that is similar in spirit t o that used by Reingold, Shaltiel and Wigderson in order to handle sources with entropy less than $\sqrt{n}$.

I will not assume any pr evious knowledge on extractors and dispersers. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013660 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110713T130000 DTEND;TZID="Asia/Jerusalem":20110713T150000 DTSTAMP;TZID="Asia/Jerusalem":20110713T130000 FREEBUSY;FBTYPE=BUSY:20110713T130000/20110713T150000 SUMMARY;LANGUAGE=en-US:The msc talk by Dmitry Pidan about Selective Predic tion of Financial Trends with Hidden Markov Models at 2011-07-13 13:00:00 DESCRIPTION;LANGUAGE=en-US:Focusing on short term trend prediction in a fi nancial context we consider the problem of selective prediction whereby the predictor can abstain from prediction in order to improve its perform ance. The main characteristic of selective predictors is the trade-off they exhibit between error and coverage rates. In the context of classific ation selective prediction is termed "classification with a reject opti on", and there the main idea for implementing rejection is Chow's ambiguit y principle In this talk we examine two types of selective HMM predictor s. The first is an ambiguity-based rejection in the spirit of Chow. The se cond is a specialized mechanism for HMMs that identifies low quality HMM states and abstain from prediction in those states. We call this model selective HMM (sHMM). In both approaches we can trade-off prediction cover age to gain better accuracy in a controlled manner. We compare the perfo rmance of ambiguity-based HMM rejection technique to that of the sHMM appr oach, demonstrate the effectiveness of both methods and the superiority of the sHMM model. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013590 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110713T133000 DTEND;TZID="Asia/Jerusalem":20110713T153000 DTSTAMP;TZID="Asia/Jerusalem":20110713T133000 FREEBUSY;FBTYPE=BUSY:20110713T133000/20110713T153000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Itsik Pe'er (Compu ter Science,Columbia University) about Bioinformatics Forum: Identity by D escent in Medical and Population Genetics at 2011-07-13 13:30:00 DESCRIPTION;LANGUAGE=en-US:Abstract: Shared segments that are Identical-by -Descent (IBD) from a recent common ancestor of purported unrelateds pro vide unique source of information for population and medical genetics: I BD addresses recent, rare variation, and is therefore key for interpreti ng data on the seam between SNP-array and sequencing studies. We have de veloped a rapid algorithm to efficiently detect IBD segments, enabling s uch analysis in large cohorts. We formalize this analysis in the context of a likelihood model, and show the utility of IBD detection in exposin g fine population structure and recent demopographic events with precisi on and dynamic range unavailable previously. We present results on simul ated data as well as examples from East Africans, who demonstrate genetics consistent with tribal demography, and from Ashkenazi Jews, who are uni que in having expanded rapidly from a bottleneck population of hundreds to a population of millions. In terms of medical applications, we develo p a method that leverages IBD analysis for testing association to rare a lleles. We apply this to SNP- and sequencing-data in isolated and genera l populations. We detect associated variants in publicly available data that had been reported by subsequent, larger studies, and replicate asso ciation signals from our own data. Combining both medical- and populatio n-genetics aspects, we show how IBD enables affordable population-based sequencing.

As an appendix, I will tell the story of the firs t animal mutant to be mapped by whole genome sequencing. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400013610 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110713T150000 DTEND;TZID="Asia/Jerusalem":20110713T170000 DTSTAMP;TZID="Asia/Jerusalem":20110713T150000 FREEBUSY;FBTYPE=BUSY:20110713T150000/20110713T170000 SUMMARY;LANGUAGE=en-US:The phd talk by Anya Levin about Handovers with For ward Admission Control for Adaptive TCP Streaming in Multihop Wireless Net works at 2011-07-13 15:00:00 DESCRIPTION;LANGUAGE=en-US:Media streaming over TCP is becoming popular be cause TCP's congestion control provides remarkable stability to the Intern et. However, TCP also introduces significant latency and throughput variab ility in the presence of mobility and frequent handovers. The effect of ha ndovers can be mitigated if the packets received by the old base station d uring handover are forwarded to the new base station. When the bandwidth o f the backbone is scarce, as in a multihop wireless network, packet forwar ding is limited. In such a case, the base stations must be selective about which packets to forward and which to drop. In this paper we define this as an optimization problem, propose algorithms for solving it, and validat e the effectiveness of these algorithms using analysis and simulations. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013640 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110720T120000 DTEND;TZID="Asia/Jerusalem":20110720T140000 DTSTAMP;TZID="Asia/Jerusalem":20110720T120000 FREEBUSY;FBTYPE=BUSY:20110720T120000/20110720T140000 SUMMARY;LANGUAGE=en-US:The msc talk by Omer Weissbrod about Linkage Analys is in the Presence of Germline Mosaicism at 2011-07-20 12:00:00 DESCRIPTION;LANGUAGE=en-US:Genetic linkage analysis is a widely used stati stical method for associating disease genes with their location on the chr omosome. This method has proved to be very successful in mapping genes inv olved in simple Mendelian diseases. However, it is less powerful in mappin g genes that do not follow Mendelian inheritance patterns. With the growin g availability of genomic data, it becomes increasingly clear that some hu man genetic traits do not follow such patterns, and thus their associated genes may elude detection. This finding motivates the development of new s tatistical tests that are more suitable to genetic mapping of such non-sta ndard genetic traits. In this work we consider a genetic condition call ed germline mosaicism that does not follow standard Mendelian inheritance patterns. Germline mosaicism violates some of the assumptions that underli e standard genetic linkage analysis. The standard method of genetic linkag e analysis is therefore not suitable for analysis of traits caused by this condition. We develop a statistical model that is suitable for analysis o f traits caused by this condition, and develop a statistical test to deter mine whether a genetic trait has been introduced by germline mosaicism. We also extend the method of genetic linkage analysis to incorporate germlin e mosaicism. Finally, we use our extended statistical model to provide sol id statistical evidence for the location of the gene responsible for a syn drome recently discovered in Israel, called the MDN syndrome. This finding may prompt further genetic research to help identify exactly the causativ e gene of this syndrome. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013650 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110720T133000 DTEND;TZID="Asia/Jerusalem":20110720T153000 DTSTAMP;TZID="Asia/Jerusalem":20110720T133000 FREEBUSY;FBTYPE=BUSY:20110720T133000/20110720T153000 SUMMARY;LANGUAGE=en-US:The colloq_cs_ee talk by Alon Orlitsky (ECE and CSE , UC San Diego) about Colloq_CS_EE: String Reconstruction from Substring C ompositions at 2011-07-20 13:30:00 DESCRIPTION;LANGUAGE=en-US:Motivated by mass-spectrometry protein sequenci ng, we consider the simple problem of reconstructing a string from its s ubstring compositions. Relating the question to the long-standing turnpi ke problem, polynomial factorization, and cyclotomic polynomials, we cle anly characterize the lengths of reconstructable strings and the structu re of non-reconstructable ones. The talk is elementary and self containe d and covers work with Jayadev Acharya, Hirakendu Das, Olgica Milenkovic , and Shengjun Pan. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1003 UID:111ec4410400013700 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110721T143000 DTEND;TZID="Asia/Jerusalem":20110721T163000 DTSTAMP;TZID="Asia/Jerusalem":20110721T143000 FREEBUSY;FBTYPE=BUSY:20110721T143000/20110721T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Avinatan Hassidim SPECIAL GUEST LECTURE about Finding a job and the two body problem at 2011-07-21 14:30:0 0 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013710 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110725T183000 DTEND;TZID="Asia/Jerusalem":20110725T203000 DTSTAMP;TZID="Asia/Jerusalem":20110725T183000 FREEBUSY;FBTYPE=BUSY:20110725T183000/20110725T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Tomer Ashur about H aifux Club: How to Spread Knowledge Throughout the World While Wearing On ly Your Slippers (or Wikimedia, Wikipedia and free content projects) at 2011-07-25 18:30:00 DESCRIPTION;LANGUAGE=en-US:Since its first emergence in 2001, Wikipedia ha d grown drastically to become the fifth most viewed website over the Inter net in 2012. With over 12,000,000 articles in more than 250 languages th is is now the largest source of information ever existed. The Wikimedia foundation has been founded in 2003 as a non-profit organization to supp ort Wikipedia as well as other online and offline free content projects. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400013630 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110728T113000 DTEND;TZID="Asia/Jerusalem":20110728T133000 DTSTAMP;TZID="Asia/Jerusalem":20110728T113000 FREEBUSY;FBTYPE=BUSY:20110728T113000/20110728T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Shachar Shem-Tov (CS, Techni on) about Pixel Club: Topics in Over-parametrized based Variational Method s at 2011-07-28 11:30:00 DESCRIPTION;LANGUAGE=en-US:We discuss a variational methodology, which inv olves locally modeling of data from noisy samples, combined with global model parameter regularization. We show that this methodology encompass es many previously proposed algorithms, from the celebrated moving leas t squares methods to the globally optimal over-parametrization methods r ecently published for smoothing and optic flow estimation.

Ho wever, the unified look at the range of problems and methods previously considered also suggests a wealth of novel global functionals and local modeling possibilities.

Specifically, we show that a new non-lo cal variational functional provided by this methodology greatly improves robustness and accuracy in local model recovery compared to previous me thods.

The proposed methodology may be viewed as a basis for a general framework for addressing a variety of common problem domains in signal and image processing and analysis, such as denoising, adaptive sm oothing, reconstruction and segmentation. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013670 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110802T113000 DTEND;TZID="Asia/Jerusalem":20110802T133000 DTSTAMP;TZID="Asia/Jerusalem":20110802T113000 FREEBUSY;FBTYPE=BUSY:20110802T113000/20110802T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Gilead Tadmor (Northeastern University, Boston) about Pixel Club: Modeling Fluid Flow on Inertial Mani folds: Physics, Geometry and the Challenge of Model Reduction at 201 1-08-02 11:30:00 DESCRIPTION;LANGUAGE=en-US:Model order reduction is essential for feasible analysis, design, realtime control of distributed systems. Recent uses also include accelerating detailed simulations and the extraction of act ionable meaning from large scale data streams. Alas, a mature and math ematically rigorous theory is largely limited to the linear case, and ev en there, the mere computational complexity its tools entail, restrict i ts use to relatively moderate dimensions. Heuristics fill in the gap, wi th successions of intuitive patches, with very mixed results.

Our motivation comes from active flow control (AFC), a field driven by n eeds of truly epic proportions: From energy efficiency of air, ground an d maritime transportation, through clean and efficient combustion and wi nd energy, micro-fluidic, diagnostic chips, artificial hearts and other bio-engineering designs, to HVAC and bio-chem defense. Arguably, the la ck of adequate low order models is a major impediment for real progress and transition to wide-range usage.

We present a rational a pproach to Galerkin-type model order reduction that tracks and addresses longstanding impediments at their first-principles roots - the Navier-S tokes equation. Embedding flow states in parameterized inertial manifold s, is used as to address difficult practical and conceptual problems, in cluding unsteady boundaries (e.g., flapping wings), boundary actuation and transient deformation of coherent structures. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013720 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110815T183000 DTEND;TZID="Asia/Jerusalem":20110815T203000 DTSTAMP;TZID="Asia/Jerusalem":20110815T183000 FREEBUSY;FBTYPE=BUSY:20110815T183000/20110815T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Aviv Sharon about H aifux Club: 0 A.D. Revisited (And Perhaps a Few Words About Wikimania 2011 ) at 2011-08-15 18:30:00 DESCRIPTION;LANGUAGE=en-US:0 A.D. is a FOSS game of ancient warfare, belon ging to a genre of games called Real-Time Strategy (RTS). It is mostly imp lemented in C++, along with scripts in JavaScript, and runs on Windows, Li nux and Mac OS. The game was last presented at Haifux in December 2009 and has developed tremendously since then, with advances in graphics, A* path finding, opponent AI and more. Aviv will demonstrate some of the new featu res of the game of particular interest to Comp Sci people and explain how he manages the PR and social media aspect of a volunteer FOSS project. Pro bable bonus: Pictures and experiences from Wikimania 2011, where Aviv is s et to assist the organizers with media and PR aspects. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400013730 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110817T140000 DTEND;TZID="Asia/Jerusalem":20110817T160000 DTSTAMP;TZID="Asia/Jerusalem":20110817T140000 FREEBUSY;FBTYPE=BUSY:20110817T140000/20110817T160000 SUMMARY;LANGUAGE=en-US:The msc talk by David Yanay about Supervised Learni ng of Semantic Relatedness at 2011-08-17 14:00:00 DESCRIPTION;LANGUAGE=en-US:We propose and study a novel supervised approac h to learning semantic relatedness from examples. Using an empirical risk minimization approach our algorithm computes a weighted measure of te rm co-occurrence with respect to a corpus of text documents, and utilize s the labeled examples to fit the model to the training sample. Our meth od is corpus independent and can essentially rely on any sufficiently lar ge (unstructured) collection of coherent texts. We present the results of a range of experiments from large to small scale. These results indicat e that the proposed method is effective and competitive with the state-o f-the-art. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013690 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110824T130000 DTEND;TZID="Asia/Jerusalem":20110824T150000 DTSTAMP;TZID="Asia/Jerusalem":20110824T130000 FREEBUSY;FBTYPE=BUSY:20110824T130000/20110824T150000 SUMMARY;LANGUAGE=en-US:The msc talk by Anat Hashavit about An Unbiased Rat ional Decision Making Procedure for Multiple-Adversary Environments at 201 1-08-24 13:00:00 DESCRIPTION;LANGUAGE=en-US:In binary-utility games, an agent can have only two possible utility values for final states, 1 (win) and 0 (lose). We define an unbiased rational agent as one that seeks to maximize its util ity value, but is equally likely to choose between states with the same utility value. In particular, it will prefer winning over losing but will be indifferent as to which winning ( or losing state) is chosen. This in duces a probability distribution over the game tree, from which an agent can infer its probability to win. A single adversary binary game is one where there are only two possible outcomes, so that the winning probabili ties remain binary values. In this case, the rational action for an agen t is to play minimax. In this work we focus on the more complex, multipl e-adversary environment, where an agent is met with at least two adversa ries. We propose a new algorithmic framework where agents try to maximiz e their winning probabilities. We begin by theoretically analyzing why a n unbiased rational agent should take our approach in an unbounded environ ment and not that of the existing Paranoid or MaxN algorithms. We then e xpand our framework to a resource-bounded environment, where winning pro babilities are estimated, and show empirical results supporting our clai ms. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013750 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110829T183000 DTEND;TZID="Asia/Jerusalem":20110829T203000 DTSTAMP;TZID="Asia/Jerusalem":20110829T183000 FREEBUSY;FBTYPE=BUSY:20110829T183000/20110829T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Amir Sagie (Arig Pr oject) about Haifux Club: Mesh Networks | Hacking the T3lc0 Model at 2011 -08-29 18:30:00 DESCRIPTION;LANGUAGE=en-US:Want to build your own Telco? you'll probable n eed mesh power. Avoid past mistakes by learning about the history of mesh networks, hear how the first wi-fi router was liberated and be sure to che ckout what we're doing in project Arig ( http://arig.org.il), here in Isra el! Be sure to attend the router emancipation party afterwords: bring your wi-fi router and wash away all it's sins by flashing it with a Foss OS su ch as OpenWRT. complete redemption guaranteed. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400013790 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110913T143000 DTEND;TZID="Asia/Jerusalem":20110913T163000 DTSTAMP;TZID="Asia/Jerusalem":20110913T143000 FREEBUSY;FBTYPE=BUSY:20110913T143000/20110913T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Udi Manber SPECIAL LECTURE about Search: past, present, and some possible futures at 2011-09-13 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013780 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110925T120000 DTEND;TZID="Asia/Jerusalem":20110925T140000 DTSTAMP;TZID="Asia/Jerusalem":20110925T120000 FREEBUSY;FBTYPE=BUSY:20110925T120000/20110925T140000 SUMMARY;LANGUAGE=en-US:The msc talk by Vadim Eisenberg about Programming A pplications over the Semantic Web at 2011-09-25 12:00:00 DESCRIPTION;LANGUAGE=en-US:Two of the hardest problems of developing data- processing applications are: (1) integrating data from heterogeneous sou rces, and (2) handling the inherent discrepancies between data models of the sources and models of programming languages, e.g., the object-relat ional impedance mismatch. The Semantic Web is a set of technologies (RDF , RDFS, OWL, SPARQL) that facilitate data integration. However, it does not solve the impedance mismatch problem. It merely exchanges object-rel ational impedance mismatch with object-RDF impedance mismatch. In the talk I will illustrate a solution to the impedance mismatch problem - a hybrid of RDF and objects. Essentially, we modify an object-oriented la nguage so that RDF data items become first-class citizens of the languag e, and objects of the language become first-class citizens of RDF. The b enefits of the hybrid model and of the modified programming language are as follows: (1) it becomes natural to use the language as a persistent programming language, where persistence issues are handled implicitly; ( 2) tools from both models, (such as optimizers, syntax checkers, query a nd reasoning engines) can be applied to the artifacts of the unified mod el; and (3) programming is done over a single unified conceptual model. In the talk I will present the modified programming language, explain it s benefits and illustrate code optimizations. A prototype system will be presented. I will provide a background to the impedance mismatch probl em and to the Semantic Web. Thus, no previous knowledge of these subject s is required. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400013810 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20110926T183000 DTEND;TZID="Asia/Jerusalem":20110926T203000 DTSTAMP;TZID="Asia/Jerusalem":20110926T183000 FREEBUSY;FBTYPE=BUSY:20110926T183000/20110926T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Orna Agmon Ben-Yehu da (CS, Technion) about Haifux Club: Deconstructing Amazon EC2 Spot Instan ce Pricing at 2011-09-26 18:30:00 DESCRIPTION;LANGUAGE=en-US:Cloud providers possessing large quantities of spare capacity must either incentivize clients to purchase it or suffer lo sses. Amazon is the first cloud provider to address this challenge, by all owing clients to bid on spare capacity and by granting resources to bidder s while their bids exceed a periodically changing spot price. Amazon publi cizes the spot price but does not disclose how it is determined. By ana lyzing the spot price histories of Amazon's EC2 cloud, we reverse engineer how prices are set and construct a model that generates prices consistent with existing price traces. We find that prices are usually not market-dr iven as sometimes previously assumed. Rather, they are typically generated at random from within a tight price interval via a dynamic hidden reserve price. Our model could help clients make informed bids, cloud providers d esign profitable systems, and researchers design pricing algorithms. Jo int work with Muli Ben-Yehuda, Assaf Schuster, Dan Tsafrir: Technion CS Te ch Report CS-2011-09. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400013800 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111010T183000 DTEND;TZID="Asia/Jerusalem":20111010T203000 DTSTAMP;TZID="Asia/Jerusalem":20111010T183000 FREEBUSY;FBTYPE=BUSY:20111010T183000/20111010T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Baruch Siach about Haifux Club: How to Participate in the Linux Kernel Development (and Why) at 2011-10-10 18:30:00 DESCRIPTION;LANGUAGE=en-US:The Linux kernel is one of the largest scale fr ee software projects. More than thousand developers contribute code to e ach kernel release. Becoming one of them is not an easy challenge. First you need to familiarize yourself with the technical side of kernel deve lopment, with its unique peculiarities. Then, you need to understand and carry out the long, and sometimes painful, process of patch submission. However, the reward for this pain is great.

This lecture i s a practical introduction to the kernel development process. I'll first explain why you (and your employer) should consider submitting patches to the mainline kernel. Then, I'll show how this process works, and what it takes to get your code into the mainline kernel. This lecture is not about the technical side of kernel development.

The intend ed audience for this lecture is kernel hackers wannabes, and those doing kernel development who do not participate (yet) in the mainline kernel development. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400013830 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111025T113000 DTEND;TZID="Asia/Jerusalem":20111025T133000 DTSTAMP;TZID="Asia/Jerusalem":20111025T113000 FREEBUSY;FBTYPE=BUSY:20111025T113000/20111025T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Erhardt Barth (Institute for Neuro- and Bioinformatics, University of Luebeck, Germany) about Pixe l Club: Gesture-based Interaction with 3D Cameras at 2011-10-25 11:30:00 DESCRIPTION;LANGUAGE=en-US:In this talk, Prof. Erhardt will discuss the AR TTS project (www.artts.eu) and summary its main results. He will also gi ve an overview of the main activities he is conducted at gestigon, a com pany he created that deals with Gesture technologies (www.gestigon.com). Prof. Erhardt will also extends the discussion on different research topics that have been applied in this context to
(i) the geometr ical approach to feature extraction,
(ii) the principle of sparse co ding, and
(iii) the pose tracker based on self-organizing maps. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400013870 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111026T113000 DTEND;TZID="Asia/Jerusalem":20111026T133000 DTSTAMP;TZID="Asia/Jerusalem":20111026T113000 FREEBUSY;FBTYPE=BUSY:20111026T113000/20111026T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Lavy Libman (University of Sydne y) about ceClub: On Distributed Coordination Strategies in Cooperative Wir eless Networks at 2011-10-26 11:30:00 DESCRIPTION;LANGUAGE=en-US:Cooperative and opportunistic communication tec hniques in wireless networks promise significant performance benefits over traditional methods that do not exploit the broadcast nature of wireless transmissions. However, such techniques generally require coordination amo ng the participating nodes, e.g. to discover available neighbors or negoti ate the best course of action after every packet broadcast. The associated coordination overheads negate much of the cooperation benefits for applic ations with strict latency requirements, or in networks with highly dynami c topologies. This talk will highlight distributed cooperation strategies, where each node overhearing a packet decides independently whether to ret ransmit it, without explicit coordination with the source, intended receiv er, or other neighbours in the vicinity. We formulate two non-traditional optimization problems that arise in this context, present their solution m ethodology and demonstrate the superior performance attained by the respec tive strategies. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1021 UID:111ec4410400013900 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111031T183000 DTEND;TZID="Asia/Jerusalem":20111031T203000 DTSTAMP;TZID="Asia/Jerusalem":20111031T183000 FREEBUSY;FBTYPE=BUSY:20111031T183000/20111031T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Abel Gordon (IBM HR L) about Haifux Club: Bare-Metal Performance for I/O Virtualization at 201 1-10-31 18:30:00 DESCRIPTION;LANGUAGE=en-US:Direct device assignment enhances the performan ce of guest virtual machines by allowing them to communicate with I/O devi ces without host involvement. But even with device assignment, guests are still unable to approach bare-metal performance, because the host intercep ts all interrupts, including those interrupts generated by assigned device s to signal to guests the completion of their I/O requests. The host invol vement induces multiple unwarranted guest/host context switches, which sig nificantly hamper the performance of I/O intensive workloads. To solve thi s problem, we present ELI (ExitLess Interrupts), a software-only approach for handling interrupts within guest virtual machines directly and securel y. By removing the host from the interrupt handling path, ELI manages to i mprove the throughput and latency of unmodified, untrusted guests by 1.3x- -1.6x, allowing them to reach 97%-100% of bare-metal performance even for the most demanding I/O-intensive workloads.

Joint work wit h Nadav Amit, Nadav Har'El, Muli Ben-Yehuda, Alex Landau, Assaf Schuster, and Dan Tsafrir ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400013850 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111101T113000 DTEND;TZID="Asia/Jerusalem":20111101T133000 DTSTAMP;TZID="Asia/Jerusalem":20111101T113000 FREEBUSY;FBTYPE=BUSY:20111101T113000/20111101T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Elhanan Elboher (School of C omputer Science and Engineering The Hebrew University of Jerusalem) about Pixel Club: Efficient and Accurate Image Filtering Using Running Sums at 2 011-11-01 11:30:00 DESCRIPTION;LANGUAGE=en-US:Non uniform filtering is important for many ima ge processing algorithms. However, for large kernel sizes it can become co mputationally expensive. In this talk we describe two efficient filtering techniques which are independent on kernel size. First we present Cosin e Integral Images (CII) which represent a large set of spatial and range f ilters, based on their frequency decomposition. We apply CII for fast comp utation of the spatial Gaussian and Gabor kernels, whose complexity is a c onstant O(1) operations per image pixel. We also improve previous constant time approximations of bilateral filtering.

I n addition we present a simple and very efficient method for Gaussian convolution using running sums along the image rows and columns. In order to approximate the Gaussian kernel we investigate relation between the error function used f or kernel approximation and the resulting L2 error on the output image. Ba sed on natural image statistics we propose a quadratic form error measure, and use it to approximate the Gaussian kernel by linear combination of co nstant functions. This results in very efficient Gaussian filtering method . Our experiments show that the proposed technique is faster than state of the art methods while preserving a similar accuracy.

Joint work with Prof. Michael Werman ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400013920 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111101T133000 DTEND;TZID="Asia/Jerusalem":20111101T153000 DTSTAMP;TZID="Asia/Jerusalem":20111101T133000 FREEBUSY;FBTYPE=BUSY:20111101T133000/20111101T153000 SUMMARY;LANGUAGE=en-US:The phd talk by Eyal Rozenberg about On the usefuln ess of blowing things up (combinatorially) at 2011-11-01 13:30:00 DESCRIPTION;LANGUAGE=en-US:This talk will overview the results of my docto ral research in Property Testing of dense combinatorial structures. In Pr operty Testing, we are concerned with the number of queries one has to m ake, or information one has to read, from an input combinatorial structur e in order to make a rough distinctions between 'good' and 'significantl y bad' inputs, where bad inputs are far from being good; specifically, m ost studies concerns such testing algorithms which read only a small frac tion of their input, sometimes only a constant-size amount, independentl y of the size of the input. In much of my work I have made use of 'b lowups' of smaller combinatorial objects into larger ones, analyzing the behavior of property testing algorithms applied to such blowups as input . The talk will highlight the use of blowups and aspects of their analys is, outlining how they can "force the hand" of property testing algorith ms, leading to hardness results, and somewhat surprisingly also to a meth od of strengthening desirable features of such testing algorithms. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400011970 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111101T143000 DTEND;TZID="Asia/Jerusalem":20111101T163000 DTSTAMP;TZID="Asia/Jerusalem":20111101T143000 FREEBUSY;FBTYPE=BUSY:20111101T143000/20111101T163000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Michael Kazhdan (John Hopkin s University) about Pixel Club: Fast Poisson Solvers for Signal Processing on Meshes at 2011-11-01 14:30:00 DESCRIPTION;LANGUAGE=en-US:In this talk, we will describe a new, octree-ba sed, FEM solver for performing geometry-aware signal processing on meshe s. We show that by considering the restriction of functions defined in 3 D to the surface, we can define a regular function space on the mesh tha t supports both multigrid solvers, and parallel and streaming computatio n. We will discuss applications of the solver to a number of traditional challenges, including texture stitching, parameterization, interactive geometry processing, and surface flow. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400013940 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111102T123000 DTEND;TZID="Asia/Jerusalem":20111102T143000 DTSTAMP;TZID="Asia/Jerusalem":20111102T123000 FREEBUSY;FBTYPE=BUSY:20111102T123000/20111102T143000 SUMMARY;LANGUAGE=en-US:The msc talk by Yohay Kaplan about Constant rate LD C's over a small(er) alphabet via tensored AG codes at 2011-11-02 12:30:00 DESCRIPTION;LANGUAGE=en-US:Locally decodable codes are codes that allow de coding of single message bits by reading a sublinear amount of the codewor d. These codes have been the object of intense study, of particular intere st is the trade-off between the rate of these codes and the query complexi ty of their local decoding algorithms. When limiting our attention to code s of constant rate, we know of only two such families of codes: Reed-Mulle r codes and multiplicity codes. While multiplicity codes achieve a higher rate (RM codes have rate at most 1/2), RM codes achieve better rates when we consider long codes (codes that have a block length greater then the fi eld size). In this work we introduce a new family of codes, tensored AG codes, and prove a Schwartz-Zippel type lemma to establish their parameter s. In addition, we prove that tensored hermitian codes, a particular type of tensored AG codes, are locally correctable (and thus, decodable) with p arameters comparable to RM codes, but over smaller fields. Our decoding al gorithm introduces a novel way to use the automorphisms of the hermitian f unction field. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400013960 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111102T123000 DTEND;TZID="Asia/Jerusalem":20111102T143000 DTSTAMP;TZID="Asia/Jerusalem":20111102T123000 FREEBUSY;FBTYPE=BUSY:20111102T123000/20111102T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Yohay Kaplan (CS, Techni on) about Theory Seminar: Constant rate LDC's over a small(er) alphabet vi a tensored AG codes at 2011-11-02 12:30:00 DESCRIPTION;LANGUAGE=en-US:Locally decodable codes are codes that allow de coding of single message bits by reading a sublinear amount of the codewor d. These codes have been the object of intense study, of particular intere st is the trade-off between the rate of these codes and the query complexi ty of their local decoding algorithms. When limiting our attention to code s of constant rate, we know of only two such families of codes: Reed-Mulle r codes and multiplicity codes. While multiplicity codes achieve a higher rate (RM codes have rate at most 1/2), RM codes achieve better rates when we consider long codes (codes that have a block length greater then the fi eld size).

In this work we introduce a new family of codes, t ensored AG codes, and prove a Schwartz-Zippel type lemma to establish thei r parameters. In addition, we prove that tensored hermitian codes, a parti cular type of tensored AG codes, are locally correctable (and thus, decoda ble) with parameters comparable to RM codes, but over smaller fields. Our decoding algorithm introduces a novel way to use the automorphisms of the hermitian function field. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013970 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111106T143000 DTEND;TZID="Asia/Jerusalem":20111106T163000 DTSTAMP;TZID="Asia/Jerusalem":20111106T143000 FREEBUSY;FBTYPE=BUSY:20111106T143000/20111106T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Bill Gasarch about When can you color a grid and not have any monochromatic rectangles? at 2011-11-06 14:3 0:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013820 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111108T143000 DTEND;TZID="Asia/Jerusalem":20111108T163000 DTSTAMP;TZID="Asia/Jerusalem":20111108T143000 FREEBUSY;FBTYPE=BUSY:20111108T143000/20111108T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Leonid Reyzin about Improving Cr yptography by Studying Entropy at 2011-11-08 14:30:00 DESCRIPTION;LANGUAGE=en-US:There are many different notions of information -theoretic entropy and its computational analogues. The right notion and a toolbox of lemmas can make for beautifully simple proofs. Drawing on examples from information-theoretic key agreement, leakage-resilient cr yptography, and deterministic encryption (no background in these topics is assumed), I will show how various extensions of entropy can lead to i mproved cryptographic constructions. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013990 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111109T123000 DTEND;TZID="Asia/Jerusalem":20111109T143000 DTSTAMP;TZID="Asia/Jerusalem":20111109T123000 FREEBUSY;FBTYPE=BUSY:20111109T123000/20111109T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Omer Tamuz (Weizmann Ins titute) about Theory Seminar: Rationality, Efficiency and Agreement in Mar kets and in Social Networks at 2011-11-09 12:30:00 DESCRIPTION;LANGUAGE=en-US:We will discuss some models of interacting econ omic agents on social networks. The first part of the talk will include a short introduction to the topic, including:
- Why assume agents are rational? What does it mean to be rational?
- When is it computa tionally feasible to be rational?
- When does interaction eventually lead to agreement, and when can disagreement persist indefinitely?
- When does interaction lead to efficient aggregation of information, and when is information lost?

In the second part of the talk we will describe recent results showing that the phenomena of agreement and efficiency are related in a wide spectrum of models. In particular, under different conditions, we will show that agents that reach consensu s are bound to learn from each other in the process.

Joint work with Elchanan Mossel and Allan Sly ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400014000 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111113T143000 DTEND;TZID="Asia/Jerusalem":20111113T163000 DTSTAMP;TZID="Asia/Jerusalem":20111113T143000 FREEBUSY;FBTYPE=BUSY:20111113T143000/20111113T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Noa Agmon about Multi-Robot Patr ol: From Theory to Reality at 2011-11-13 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013860 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111114T183000 DTEND;TZID="Asia/Jerusalem":20111114T203000 DTSTAMP;TZID="Asia/Jerusalem":20111114T183000 FREEBUSY;FBTYPE=BUSY:20111114T183000/20111114T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Ofer Rosenberg (AMD ) about Haifux Club: GPGPU - Motivation and Architecture (Part 1 out of 4 talks series) at 2011-11-14 18:30:00 DESCRIPTION;LANGUAGE=en-US:This is a 4 series of 4 talks about GPGPUS, int ended for the practical engineer:
1. Motivation, AMD's architecture
2. OpenCL
3.Case studies, Dos and Don'ts
4.Tools and Profiling for Performance

General Purpose GPU programming be came a hot topic in the last few years, ranging from academic studies to b eing used by commercial software products. As an example, three out of the world's top10 supercomputers (June2011 list) contain GPUs in them. This s eries of lectures focuses on OpenCL, the open standard for parallel progra mming of heterogeneous systems. The third and fourth meetings will incl ude hands-on experience, under Ofer's guidance. For these meetings, each p articipant will have to bring a laptop with:
1. Linux (this is Haif ux, the Haifa Linux Club)
2. An x86 processor, either AMD or Intel. An IPAD is not good for this purpose.
3. AMD's OpenCL implementatio n installed. It works both on AMD and Intel, installation details for thos e who need them will follow.
4. Eclipse installed.
5. If poss ible, an AMD GPU, with a version number of 54 or higher. Use "lspci | grep VGA" to find the GPU vendor & model.

To arrange for machine acc ess for those who do not have a suitable laptop, as well as prepare for th e series, please email in private to webmaster@haifux.org
• If you intend to attend the seri es
• If you have (or do not have) the required gear. In particular , if you have an AMD GPU.

Syllabus: Introduction to GPGPU Programming
General Purpose GPU programming became a hot top ic in the last few years, ranging from academic studies to being used by c ommercial software products. As an example, three out of the world's top10 supercomputers (June2011 list) contain GPUs in them. This series of lectu res focuses on OpenCL, the open standard for parallel programming of heter ogeneous systems. 1. GPGPU introduction

The first lect ure is an introduction to GPU architecture and GPGPU programing. It covers the differences between GPU and CPU architectures, and how these differen ces impose restrictions on programming GPUs. We will also touch the issue of memory aspects of GPU architecture and the overall system (CPU & GPU)

2. OpenCL overview

From the Khronos website: "O penCL" is the first open, royalty-free standard for cross-platform, parall el programming of modern processors found in personal computers, servers a nd handheld/embedded devices. This lecture will provide an overview of Ope nCL, covering the API programming aspects (such as OpenCL objects, context s, queues, events, etc.) as well as the language enhancements (such as vec tors, images, samplers, built-in functions etc.)

3. OpenCL D o's and Don'ts

This lecture provides a practical guide for p rogramming in OpenCL by doing a hands-on guided experience of writing Open CL applications and kernels. Starting from basic examples through more com plex scenarios, we will provide some tips for writing code that provides t he required correct results. We will also provide some performance tips.

4. OpenCL Optimization & Profiling

This lecture focuses on performance aspects of OpenCL. We will provide a hands-on expe rience of improving performance of OpenCL kernels by optimizing a specific example. In addition we will show ways to profile the kernel, including w orking with profiling tools such as AMD kernel profiler and gDebugger. Not e that some of issues presented in this lecture will be possible only on A MD GPUs.

More details about GPGPU Programming Lecture Series and abouth Haifux in: http://haifux.org/

ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400013980 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111116T113000 DTEND;TZID="Asia/Jerusalem":20111116T133000 DTSTAMP;TZID="Asia/Jerusalem":20111116T113000 FREEBUSY;FBTYPE=BUSY:20111116T113000/20111116T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Dror Feitelson (CS, Hebrew Univ ersity of Jerusalem) about ceClub: Observations on Linux Development at 20 11-11-16 11:30:00 DESCRIPTION;LANGUAGE=en-US:Linux is used extensively in systems research a s a platform for the implementation of new ideas, exploiting its open-sour ce nature. But Linux is also interesting as an object of study about softw are engineering. In particular, Linux defies common management theories, a s it lacks any coherent plan or management structure, but still grows at a n ever-increasing rate, while also gaining market share. We will review so me previous studies of Linux development and add new observations regardin g the lifecycle model and code complexity.

The presentation includes results of my students Ayelet Israeli, Ahmad Jbara, and Adam Mata n. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 861 UID:111ec4410400013910 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111116T123000 DTEND;TZID="Asia/Jerusalem":20111116T143000 DTSTAMP;TZID="Asia/Jerusalem":20111116T123000 FREEBUSY;FBTYPE=BUSY:20111116T123000/20111116T143000 SUMMARY;LANGUAGE=en-US:The msc talk by Yonatan Goldhirsh about Property Te sting in the Massively Parametrized Model at 2011-11-16 12:30:00 DESCRIPTION;LANGUAGE=en-US:We say that a colored ordered rooted tree H is a topological subtree of a colored ordered rooted tree T if we can map the vertices of H to vertices of T in a mapping which is one-to-one, preserve s colors, preserves the ``descendant of'' relation, preserves right-to-lef t order, and maps least common ancestors to least common ancestors. Fix an ordered rooted tree T and let F be a family of ``forbidden'' colored orde red trees. We present an algorithm for testing whether a coloring of T is free from F or is epsilon-far from being free from F, with query comp lexity depending only on F and epsilon. This is an instance of the massive ly parameterized testing model, as T itself is immutable. Families of forb idden topological subtrees generalize previously considered tree coloring properties, such as tree monotonicity [FLNRRS 02] and convexity-like prope rties [FY 07]. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400014010 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111116T143000 DTEND;TZID="Asia/Jerusalem":20111116T163000 DTSTAMP;TZID="Asia/Jerusalem":20111116T143000 FREEBUSY;FBTYPE=BUSY:20111116T143000/20111116T163000 SUMMARY;LANGUAGE=en-US:The CSpecial Talk talk by Richard M. Karp (Universi ty of California at Berkeley and International Computer Science Institute) about CSpecial Talk: Effective Heuristics for NP-Hard Problems at 2011 -11-16 14:30:00 DESCRIPTION;LANGUAGE=en-US:In many practical situations heuristic algorith ms reliably give satisfactory solutions to real-life instances of optim ization problems, despite evidence from computational complexity theory that the problems are intractable in general.Our long-term goal is to contribute to an understanding of this seeming contradiction, and to pu t the construction of heuristic algorithms on a firmer footing.
< br> As a step in this direction we describe the evolution of a succesful heuristic algorithm By Erick Moreno Centeno and the speaker for the following Global Genome Alignment problem: given the genomes of several species, an anchor pair is a pair of substrings from two different g enomes which appear to be descended from a common ancestral sequence. Given a collection of anchor pairs, construct a multiple alignment maxi mizing the number of anchor pairs that are aligned against each other. Such an alignment exhibits the common evolutionary ancestry of the set of species.

We then use the Global Genome Alignment problem t o illustrate a general approach for tuning the parameters and design ch oices within a given heuristic algorithmic strategy, assuming the avail ability of a training set of typical problem instances. This approach l eads to a decision-theoretic problem related to the Multi-Armed Bandit Problem. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013890 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111116T163000 DTEND;TZID="Asia/Jerusalem":20111116T183000 DTSTAMP;TZID="Asia/Jerusalem":20111116T163000 FREEBUSY;FBTYPE=BUSY:20111116T163000/20111116T183000 SUMMARY;LANGUAGE=en-US:The msc talk by Uri Verner about Processing Real-ti me Data Streams on Accelerator-based Systems at 2011-11-16 16:30:00 DESCRIPTION;LANGUAGE=en-US:Data stream processing applications such as sto ck exchange data analysis, VoIP streaming, and sensor data processing pose two conflicting goals: limited per-stream latency -- to satisfy the milli seconds-long hard real-time constraints of each stream, and high throughpu t -- to enable efficient processing of as many streams as possible. High-t hroughput programmable accelerators such as modern GPUs hold high potentia l to speed up the computations. However their use for hard-real time strea m processing is complicated by slow communications with CPUs, variable thr oughput changing non-linearly with the input size, and weak consistency of their local memory with respect to CPU accesses. Furthermore, their coars e grain hardware scheduler renders them unsuitable for unbalanced multi-st ream workloads. We present a general, efficient and practical algorithm for hard real-time stream scheduling in heterogeneous systems. The algori thm assigns incoming streams of different rates and deadlines to CPUs and accelerators. By employing novel stream schedulability criteria for accele rators, the algorithm finds the assignment which simultaneously satisfies the aggregate throughput requirements of all the streams and the deadline constraint of each stream alone. We use this algorithm to implement a f ramework for processing multiple data streams on GPU-CPU hybrid systems. T o the best of our knowledge, it is the first application of GPUs to hard r eal-time computations. Extensive experiments using the AES-CBC encryption kernel on thousands of streams with realistic distribution of rates and de adlines show that our framework outperforms the alternative methods by all owing 50% more streams to be processed with provably deadline-compliant ex ecution even for deadlines as short as tens of milliseconds. Overall, the combined GPU-CPU execution allows for up to 4-fold throughput increase ove r highly-optimized multi-threaded CPU-only implementation. Currently, we are working on adding support for multi-GPU systems and dynamic streams , enhancing the performance model and providing a solid theoretical backgr ound. There will be a quick introduction to GPU computing at the beginn ing. This is a direct PhD seminar. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400014040 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111117T153000 DTEND;TZID="Asia/Jerusalem":20111117T173000 DTSTAMP;TZID="Asia/Jerusalem":20111117T153000 FREEBUSY;FBTYPE=BUSY:20111117T153000/20111117T173000 SUMMARY;LANGUAGE=en-US:The CSpecial Talk talk by Ronny Lempel (Yahoo! Dire ctor of Research) about CSpecial Talk: Beyond 10 Blue Links: How Search En gines Help Users" at 2011-11-17 15:30:00 DESCRIPTION;LANGUAGE=en-US:This talk highlights some of the tools made ava ilable by search engines to aid users in formulating information needs, di gesting complex information spaces and completing tasks online. Much of th e innovation and competition in the Web search industry focuses around suc h tooling, as search engines go well beyond the concept of returning "10 b lue links" to users. Examples from multiple search engines will be shown. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013950 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111117T160000 DTEND;TZID="Asia/Jerusalem":20111117T180000 DTSTAMP;TZID="Asia/Jerusalem":20111117T160000 FREEBUSY;FBTYPE=BUSY:20111117T160000/20111117T180000 SUMMARY;LANGUAGE=en-US:The CSpecial Talk talk by Richard M. Karp (Universi ty of California at Berkeley and International Computer Science Institute) about CSpecial Talk: Theory of Computation as a Lens on the Sciences a t 2011-11-17 16:00:00 DESCRIPTION;LANGUAGE=en-US:Many processes in the physical, biological, eng ineering and social sciences involve information processing at a fundament al level and can be studied through computational models. This talk wil l describe the impact of such models on areas such as quantum computing, dtatistical physics, economics and game theory, mathematics and computa tional biology. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013880 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111120T143000 DTEND;TZID="Asia/Jerusalem":20111120T163000 DTSTAMP;TZID="Asia/Jerusalem":20111120T143000 FREEBUSY;FBTYPE=BUSY:20111120T143000/20111120T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Reut Tsarfaty about Statistical Parsing in the Face of Language Diversity at 2011-11-20 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014020 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111122T113000 DTEND;TZID="Asia/Jerusalem":20111122T133000 DTSTAMP;TZID="Asia/Jerusalem":20111122T113000 FREEBUSY;FBTYPE=BUSY:20111122T113000/20111122T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Todd Zickler (Harvard School of Engineering and Applied Sciences) about Pixel Club: Toward Computer Vision on a Tight Budget at 2011-11-22 11:30:00 DESCRIPTION;LANGUAGE=en-US:At least in the near term, micro-scale platform s like micro air vehicles and micro sensor nodes are unlikely to have po wer, volume, or mass budgets to support conventional imaging and post-ca pture processing for visual tasks like detection and tracking. These bud gets are severe enough that even common computations, such as large matr ix manipulations and convolutions, are difficult or impossible. To help overcome this, we are considering sensor designs that allow some compone nts of scene analysis to happen optically, before light strikes the sens or. I will present and analyze one class of designs in this talk. These sensors reduce power requirements through template-based optical convolu tion, and they enable a wide field-of-view within a small form. I will d escribe the trade-offs between field-of-view, volume, and mass in these sensors, and I will describe our initial efforts toward choosing effecti ve templates. I will also show examples of milli-scale prototypes for si mple computer vision tasks such as locating edges, tracking targets, and detecting faces. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400014080 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111122T143000 DTEND;TZID="Asia/Jerusalem":20111122T163000 DTSTAMP;TZID="Asia/Jerusalem":20111122T143000 FREEBUSY;FBTYPE=BUSY:20111122T143000/20111122T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by David F. Bacon about Liquid Meta l: Programming in the Age of Heterogeneous Machines at 2011-11-22 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400013840 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111122T143000 DTEND;TZID="Asia/Jerusalem":20111122T163000 DTSTAMP;TZID="Asia/Jerusalem":20111122T143000 FREEBUSY;FBTYPE=BUSY:20111122T143000/20111122T163000 SUMMARY;LANGUAGE=en-US:The CSpecial Talk talk by Ronen Bar-Nahor (Agilespa rks) about CSpecial Talk: Software Development by the Agile System at 2011 -11-22 14:30:00 DESCRIPTION;LANGUAGE=en-US:For abstract please see Hebrew page. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Amado 233 UID:111ec4410400014050 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111123T113000 DTEND;TZID="Asia/Jerusalem":20111123T133000 DTSTAMP;TZID="Asia/Jerusalem":20111123T113000 FREEBUSY;FBTYPE=BUSY:20111123T113000/20111123T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Olga Brukman (CS, Technion) abou t ceClub: Self-stabilizing Autonomic Recoverers at 2011-11-23 11:30:00 DESCRIPTION;LANGUAGE=en-US:This talk introduces theoretical foundations fo r system architectures and algorithms for creating truly robust autonomic systems -- systems that are able to recover automatically from unexpected failures. We consider various settings of system transparency. We consider black box and transparent box software packages. The general assumption i s that a software package fails when it encounters an unexpected environme nt state -- a state the package was not programmed to cope with. Creating a system that anticipates every possible environment state is not feasible due to the size of the environment. Thus, an autonomic system design shou ld imply that a system is able to overcome an unexpected environment state either by executing a recovery action that restores a legal state or by f inding a new program that respects the specifications and achieves the sof tware package goals in the current environment.

In the first part of this talk, we consider software packages to be black boxes. We pr opose modeling software package flaws (bugs) by assuming eventual Byzantin e behavior of the package. A general, yet practical, framework and paradig m for the monitoring and recovery of systems called autonomic recoverer is proposed. In the second part we consider a software package to be a trans parent box and introduce the recovery oriented programming paradigm. Progr ams designed according to the recovery oriented programming paradigm inclu de important safety and liveness properties and recovery actions as an int egral part of the program. We design a pre-compiler that produces augmente d code for monitoring the properties and executing the recovery actions up on a property violation. Finally, in the third part, we consider a highly dynamic environment, which typically implies that there are no realizable specifications for the environment, i.e., there does not exist a program t hat respects the specifications for every given environment. We suggest se arching for a program in run time by trying all possible programs on envir onment replicas in parallel. We design control search algorithms that expl oit various environment properties. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014120 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111123T120000 DTEND;TZID="Asia/Jerusalem":20111123T140000 DTSTAMP;TZID="Asia/Jerusalem":20111123T120000 FREEBUSY;FBTYPE=BUSY:20111123T120000/20111123T140000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Shai Avidan (Tel Aviv Univer sity) about Pixel Club: Coherency Sensitive Hashing at 2011-11-23 12:00:00 DESCRIPTION;LANGUAGE=en-US:Coherency Sensitive Hashing (CSH) extends Local ity Sensitivity Hashing (LSH) and PatchMatch to quickly find matching patches between two images. LSH relies on hashing, which maps similar patches to the same bin, in order to find matching patches. PatchMatch , on the other hand, relies on the observation that images are coheren t, to propagate good matches to their neighbors, in the image plane. I t uses random patch assignment to seed the initial matching. CSH relie s on hashing to seed the initial patch matching and on image coherence to propagate good matches. In addition, hashing lets it propagate inf ormation between patches with similar appearance (i.e., map to the sam e bin). This way, information is propagated much faster because it can use similarity in appearance space or neighborhood in the image plane . As a result, CSH is at least three to four times faster than PatchMa tch and more accurate, especially in textured regions, where reconstru ction artifacts are most noticeable to the human eye. We verified CSH on a new, large scale, data set of 133 image pairs.
Coherency Se nsitive Hashing (CSH) extends Locality Sensitivity Hashing (LSH) and P atchMatch to quickly find matching patches between two images. LSH rel ies on hashing, which maps similar patches to the same bin, in order t o find matching patches. PatchMatch, on the other hand, relies on the observation that images are coherent, to propagate good matches to the ir neighbors, in the image plane. It uses random patch assignment to s eed the initial matching. CSH relies on hashing to seed the initial pa tch matching and on image coherence to propagate good matches. In addi tion, hashing lets it propagate information between patches with simil ar appearance (i.e., map to the same bin). This way, information is pr opagated much faster because it can use similarity in appearance space or neighborhood in the image plane. As a result, CSH is at least thre e to four times faster than PatchMatch and more accurate, especially i n textured regions, where reconstruction artifacts are most noticeable to the human eye. We verified CSH on a new, large scale, data set of 133 image pairs.

joint work with Simon Korman. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400014090 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111123T123000 DTEND;TZID="Asia/Jerusalem":20111123T143000 DTSTAMP;TZID="Asia/Jerusalem":20111123T123000 FREEBUSY;FBTYPE=BUSY:20111123T123000/20111123T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Ron Lavi (Technion) abou t Theory Seminar: Side-communication and Efficiency of Accending Auctions at 2011-11-23 12:30:00 DESCRIPTION;LANGUAGE=en-US:We analyze the realistic, popular format of an ascending auction with anonymous item prices, when there are two items tha t are substitutes. This auction format entails increased opportunities for bidders to coordinate bids, as the bidding process is longer, and since b idders see the other bids and can respond to various signaling. This has h appened in many real auctions, e.g., in the Netherlands 3G Telecom Auction and in the FCC auctions in the US.

While on the face of it, suc h bidding behavior seems to harm economic effi ciency, we show that side-c ommunication may actually improve the social e fficiency of the auction: W e describe an ex-post sub-game perfect equilibrium, that uses limited side -communication, and is ex-post e fficient. In contrast, without side-commu nication, we show that there is no ex-post equilibrium which is ex-post e fficient in the ascending auction.

In the equilibrium strateg y we suggest, bidders start by reporting their true demands at the first s tages of the auction, and then perform a single demand reduction at a cert ain concrete point, determined using a single message exchanged between th e bidders. We show that this limited collusion opportunity resolves the st rategic problems of myopic bidding, and, quite surprisingly, improves soci al welfare instead of harming it.

Joint work with Sigal Oren . ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400014030 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111128T183000 DTEND;TZID="Asia/Jerusalem":20111128T203000 DTSTAMP;TZID="Asia/Jerusalem":20111128T183000 FREEBUSY;FBTYPE=BUSY:20111128T183000/20111128T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Ofer Rosenberg (AMD ) about Haifux Club: GPGPU - OpenCL (Part 2 out of 4 talks series) at 2011-11-28 18:30:00 DESCRIPTION;LANGUAGE=en-US:This is a 4 series of 4 talks about GPGPUS, int ended for the practical engineer:
1. Motivation, AMD's architecture
2. OpenCL
3.Case studies, Dos and Don'ts
4.Tools and Profiling for Performance

General Purpose GPU programming be came a hot topic in the last few years, ranging from academic studies to b eing used by commercial software products. As an example, three out of the world's top10 supercomputers (June2011 list) contain GPUs in them. This s eries of lectures focuses on OpenCL, the open standard for parallel progra mming of heterogeneous systems. The third and fourth meetings will incl ude hands-on experience, under Ofer's guidance. For these meetings, each p articipant will have to bring a laptop with:
1. Linux (this is Haif ux, the Haifa Linux Club)
2. An x86 processor, either AMD or Intel. An IPAD is not good for this purpose.
3. AMD's OpenCL implementatio n installed. It works both on AMD and Intel, installation details for thos e who need them will follow.
4. Eclipse installed.
5. If poss ible, an AMD GPU, with a version number of 54 or higher. Use "lspci | grep VGA" to find the GPU vendor & model.

To arrange for machine acc ess for those who do not have a suitable laptop, as well as prepare for th e series, please email in private to webmaster@haifux.org
• If you intend to attend the seri es
• If you have (or do not have) the required gear. In particular , if you have an AMD GPU.

Syllabus: Introduction to GPGPU Programming

General Purpose GPU programming became a hot t opic in the last few years, ranging from academic studies to being used by commercial software products. As an example, three out of the world 's top10 supercomputers (June2011 list) contain GPUs in them. This seri es of lectures focuses on OpenCL, the open standard for parallel progra mming of heterogeneous systems.


1. GPGPU introduction

The first lecture is an introduction to GPU architecture and GPGPU progra ming. It covers the differences between GPU and CPU architectures, and how these differences impose restrictions on programming GPUs. We will also touch the issue of memory aspects of GPU architecture and the over all system (CPU & GPU)


2. OpenCL overview

From the Khr onos website: "OpenCL" is the first open, royalty-free standard for cro ss-platform, parallel programming of modern processors found in personal computers, servers and handheld/embedded devices. This lecture will prov ide an overview of OpenCL, covering the API programming aspects (such a s OpenCL objects, contexts, queues, events, etc.) as well as the langua ge enhancements (such as vectors, images, samplers, built-in functions etc.)


3. OpenCL Do's and Don'ts

This lecture provides a practical guide for programming in OpenCL by doing a hands-on guided experience of writing OpenCL applications and kernels. Starting from ba sic examples through more complex scenarios, we will provide some tips for writing code that provides the required correct results. We will also provide some performance tips.


4. OpenCL Optimization & Pro filing

This lecture focuses on performance aspects of OpenCL. We w ill provide a hands-on experience of improving performance of OpenCL ke rnels by optimizing a specific example. In addition we will show ways t o profile the kernel, including working with profiling tools such as AM D kernel profiler and gDebugger. Note that some of issues presented in this lecture will be possible only on AMD GPUs.

Lecture slides in PDF format

Back to the Club's homepage ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400014070 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111129T113000 DTEND;TZID="Asia/Jerusalem":20111129T133000 DTSTAMP;TZID="Asia/Jerusalem":20111129T113000 FREEBUSY;FBTYPE=BUSY:20111129T113000/20111129T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Tsvi Achler (Los Alamos Nati onal Labs) about Pixel Club: Generative Reconstruction: An Efficient Way to Flexibly Store andRecognize Patterns at 2011-11-29 11:30:00 DESCRIPTION;LANGUAGE=en-US:The brain has recognition capabilities that rem ain unmatched by computer algorithms. We hypothesize that using matched feedforward-feedback connections, recognition centers of the brain re construct an internal copy of inputs using knowledge the brain has previ ously accumulated, in accordance with a class models called "generative models". Subsequently, it minimizes the error between the internal copy and the input from the environment. We study how this strategy may ena ble simple flexible learning, scalability, overcome combinatorial proble ms associated with pattern mixtures, and display cognitive phenomena. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400014180 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111129T143000 DTEND;TZID="Asia/Jerusalem":20111129T163000 DTSTAMP;TZID="Asia/Jerusalem":20111129T143000 FREEBUSY;FBTYPE=BUSY:20111129T143000/20111129T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Avi Mendelson about Back to the Future: the reborn of Dataflow computational models at 2011-11-29 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014100 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111130T113000 DTEND;TZID="Asia/Jerusalem":20111130T133000 DTSTAMP;TZID="Asia/Jerusalem":20111130T113000 FREEBUSY;FBTYPE=BUSY:20111130T113000/20111130T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Shachar Raindel (EE, Technion) a bout ceClub: Replicate and Bundle (RnB) - A Relief for Certain Data Center Bottlenecks at 2011-11-30 11:30:00 DESCRIPTION;LANGUAGE=en-US:In this talk, we present the Replicate and Bund le (RnB) scheme for relieving back-end processor and network bottlenecks i n read-mostly key-value storage systems wherein each user request spawns a large number of back-end small-item requests. This is common in Web 2.0 a nd online social network systems. Adding processors is of little help beca use this increases the number of back-end requests per user request, there by also increasing the overall processor and network load. Instead, RnB ad ds memory and replicates the stored data, thereby permitting each user req uest to be serviced through fewer back-end requests, reducing the total am ount of work required from the network-processing and storage components o f the system. The scheme is very easy to deploy, completely distributed, a nd while oblivious to the workload, beneficially exploits its spatial loca lity.

We have studied RnB through simulation, and augmented this with a micro-benchmark in order to estimate the expected actual syste m performance.

Our results show that by using RnB, two major scaling bottlenecks are considerably relieved: 1) the TCP InCast issue an d 2) the memcached multi-get hole. When utilizing the requests' locality, we achieve a considerable improvement in these bottlenecks with relatively little additional memory. In fact, RnB can use existing replicas (include d for fault tolerance), in which case the extra cost is minimal if any. Fi nally, we note that this is yet another example of judicious exploitation of redundancy for performance enhancement. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 861 UID:111ec4410400014160 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111130T123000 DTEND;TZID="Asia/Jerusalem":20111130T143000 DTSTAMP;TZID="Asia/Jerusalem":20111130T123000 FREEBUSY;FBTYPE=BUSY:20111130T123000/20111130T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Elad Hazan (Technion) ab out Theory Seminar: Efficient Optimization in Machine Learning at 2011-1 1-30 12:30:00 DESCRIPTION;LANGUAGE=en-US:Linear classification is a fundamental problem of machine learning, in which positive and negative examples of a concept are represented in Euclidean space by their feature vectors, and we seek t o find a hyperplane separating the two classes of vectors.

I n this talk we'll describe recent advances in efficient algorithms for lin ear classification and related machine learning problems. In particular we 'll describe the first sublinear-time (information-optimal) algorithms for linear classification and discuss when can optimization time meets the in formation limit. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400014170 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111130T140000 DTEND;TZID="Asia/Jerusalem":20111130T160000 DTSTAMP;TZID="Asia/Jerusalem":20111130T140000 FREEBUSY;FBTYPE=BUSY:20111130T140000/20111130T160000 SUMMARY;LANGUAGE=en-US:The phd talk by Eliyahu Osherovich about Numerical Methods for Phase Retrieval at 2011-11-30 14:00:00 DESCRIPTION;LANGUAGE=en-US:In this work we consider the problem of the rec onstruction of a signal from the magnitude of its Fourier transform, also known as phase retrieval. The problem arises in many areas of astronomy, crystallography, optics, and coherent diffraction imaging (CDI). Our ma in goal is to develop an efficient reconstruction method based on continu ous optimization techniques. Unlike current reconstruction methods, which are based on alternating projections, our approach leads to a much faste r and more robust method. However, all previous attempts to employ contin uous optimization methods, such as Newton-type algorithms, to the phase r etrieval problem failed. In this work we provide an explanation for this failure and based on this explanation we devise a sufficient condition th at allows development of new reconstruction methods---approximately known Fourier phase. We demonstrate that a rough (up to $\pi/2$ radians) Fouri er phase estimate practically guarantees successful reconstruction by any reasonable method. We also present a new reconstruction method whose re construction time is orders of magnitude faster than that of the current method-of-choice in phase retrieval---Hybrid Input-Output (HIO). Moreover , our method is capable of successful reconstruction even in the situatio ns where HIO is known to fail. We also extended our method to other appli cations: Fourier domain holography, and interferometry. Additionally w e(*) developed a new sparsity-based method for sub-wavelength CDI. Using this method we demonstrated experimental resolution exceeding several tim es the physical limit imposed by the diffraction light properties (so cal led diffraction limit). (*) The work on sub-wavelength CDI was done in collaboration with Prof. M. Segev's group from the Technion Physics Depar tment, Solid State Institute. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400014140 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111206T113000 DTEND;TZID="Asia/Jerusalem":20111206T133000 DTSTAMP;TZID="Asia/Jerusalem":20111206T113000 FREEBUSY;FBTYPE=BUSY:20111206T113000/20111206T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Maria Zontak (CS & Applied M athematics, Weizmann Institute of Science) about Pixel Club: Internal Statistics of a Single Natural Image at 2011-12-06 11:30:00 DESCRIPTION;LANGUAGE=en-US:Statistics of 'natural images' provides useful priors for solving under-constrained problems in Computer Vision. Such s tatistics is usually obtained from large collections of natural images. We claim that the substantial internal data redundancy within a single n atural image (e.g., recurrence of small image patches), gives rise to po werful internal statistics, obtained directly from the image itself. Whi le internal patch recurrence has been used in various applications, we p rovide a parametric quantification of this property. We show that the li kelihood of an image patch to recur at another image location can be exp ressed parametrically as a function of the spatial distance from the pat ch, and its gradient content. This "internal parametric prior" is used t o improve existing algorithms that rely on patch recurrence.

Moreover, we show that internal image-specific statistics is often more powerful than general external statistics, giving rise to more powerful image-specific priors. In particular:
(i) Patches tend to recur mu ch more frequently (densely) inside the same image, than in any random e xternal collection of natural images. (ii) To find an equally good exter nal representative patch for all the patches of an image, requires an ex ternal database of hundreds of natural images.
(iii) Internal stat istics often has stronger predictive power than external statistics, ind icating that it may potentially give rise to more powerful image-specifi c priors.

*Joint work with Michal Irani. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400014230 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111206T143000 DTEND;TZID="Asia/Jerusalem":20111206T163000 DTSTAMP;TZID="Asia/Jerusalem":20111206T143000 FREEBUSY;FBTYPE=BUSY:20111206T143000/20111206T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Gil Segev about From Cryptograph y to Algorithms and Back Again at 2011-12-06 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014110 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111207T113000 DTEND;TZID="Asia/Jerusalem":20111207T133000 DTSTAMP;TZID="Asia/Jerusalem":20111207T113000 FREEBUSY;FBTYPE=BUSY:20111207T113000/20111207T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Alex Kogan (Technion) about ceCl ub: Crafting Fast Wait-Free Algorithms at 2011-12-07 11:30:00 DESCRIPTION;LANGUAGE=en-US:Lock-freedom is a progress guarantee that ensur es overall program progress. Wait-freedom is a stronger progress guarantee that ensures the progress of each thread in the program. The latter prope rty is valuable for systems that need to be responsive, such as real-time systems, operating systems, etc. While many practical lock-free algorithms are known in the literature, constructing wait-free algorithms is conside red a complicated task that results in inefficient implementations. In thi s talk, we present first construction of a wait-free queue and a methodolo gy called fast-path-slow-path, which allows the wait-free algorithms (and our queue in particular) to practically match the performance of the lock- free ones. Subsequent work has used this methodology to create other effic ient wait-free algorithms.

bio: Alex Kogan is a Ph.D. studen t in the Department of Computer Science at the Technion. He holds M.Sc. (2 008) and B.A (Summa Cum Laude, 2002) degrees from the same department. His research interests lie at the intersection of distributed and mobile comp uting, and in parallel computing. Specifically, he is interested in constr ucting efficient and reliable distributed algorithms for mobile networks, utilizing emergent technologies for wireless communication and the availab ility of cloud services. Recently, he became interested in exploring and b oosting the performance of modern multi-core systems by devising efficient concurrent algorithms. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400014270 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111207T123000 DTEND;TZID="Asia/Jerusalem":20111207T143000 DTSTAMP;TZID="Asia/Jerusalem":20111207T123000 FREEBUSY;FBTYPE=BUSY:20111207T123000/20111207T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Amitabh Trehaan (Technio n) about Theory Seminar: Data structures for self-healing networks: Forgiv ingGraph and Xheal at 2011-12-07 12:30:00 DESCRIPTION;LANGUAGE=en-US:In this talk, we consider the problem of self-h ealing in reconfigurable networks (e.g. peer-to-peer and wireless mesh net works) that are under repeated attack by an omniscient adversary and propo se fully distributed algorithms that 'heal' certain global and local prope rties while doing only local changes and using only local information.

Our model assumes repeated attack by an omniscient adversary. We assume that, over a sequence of rounds, an adversary either inserts a n ode with arbitrary connections or deletes an arbitrary node from the netwo rk. The network responds to each such change by quick "repairs," which con sist of adding or deleting a small number of edges. We shall cover brie fly, either one or both of the following algorithms, as time permits:
Forgiving Graph [PODC 2009; The forgiving graph: a distributed da ta structure for low stretch under adversarial attack] preserves closenes s of nodes i.e. network stretch after adversarial deletions or insertions, without increasing node degrees by more than a constant factor. It also i ntroduces a simple but interesting mergable data structure called half-ful l tree and shows how to use it to implement the algorithm in the distribut ed setting.

Xheal [PODC 2011; Xheal: localized self-healing using expanders] maintains good expansion and spectral properties of the network, also keeping the network connected. Moreover, Xheal does this wh ile allowing only low stretch and degree increase per node. The central id ea is to reconnect nodes that have lost a neighbor by k-regular expanders while not allowing degrees to blow up. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400014250 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111207T150000 DTEND;TZID="Asia/Jerusalem":20111207T170000 DTSTAMP;TZID="Asia/Jerusalem":20111207T150000 FREEBUSY;FBTYPE=BUSY:20111207T150000/20111207T170000 SUMMARY;LANGUAGE=en-US:The msc talk by Omer Levy about Teaching Machines t o Learn by Metaphors at 2011-12-07 15:00:00 DESCRIPTION;LANGUAGE=en-US:Humans have an uncanny ability to learn new con cepts with very few examples. Cognitive theories have suggested that this is done by utilizing prior experience of related tasks. We propose to emul ate this process in machines, by transforming new problems into old ones. These transformations are called metaphors. Obviously, the learner is not given a metaphor, but must acquire one through a learning process. We show that learning metaphors yield better results than existing transfer learn ing methods. Moreover, we argue that metaphors give a qualitative assessme nt of task relatedness. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400014210 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111213T113000 DTEND;TZID="Asia/Jerusalem":20111213T133000 DTSTAMP;TZID="Asia/Jerusalem":20111213T113000 FREEBUSY;FBTYPE=BUSY:20111213T113000/20111213T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Anat Levin (CS & Mathematic s, Weizmann Institute of Science) about Pixel Club: Patch Complexity, Fin ite Pixel Correlations and Optimal Denoising at 2011-12-13 11:30:00 DESCRIPTION;LANGUAGE=en-US:Restoration tasks, such as image denoising, are ill posed problems, often solved with image priors. As image priors are only approximate, this yields suboptimal restoration results. Given t he numerous works on image priors and image denoising, it is thus import ant to understand the inherent limits posed by natural image statistics , and what potential gains we may expect from additional years of resear ch efforts. Recent studies avoided image priors with a non-parametric ap proach, but were restricted to small patches, still leaving unclear how results improve for larger patches.

In this paper we aim to understand the ``patch complexity" of natural image statistics and the potential gain from an increase in patch size.

We first co nsider the computational aspect and study the relation between performan ce gain and sample size requirements in a non parametric approach. In the second part we put computational restrictions aside, and study the i nherent statistical aspect. We show that scale invariance of natural i mages yields both a strictly positive lower bound on denoising performa nce and a power law convergence to it as a function of patch size. Extr apolating our finite support results gives a ballpark estimate of the be st achievable denoising. We also suggest directions for potential improv ements of current algorithms.

Joint work with Boaz Nadler, F redo Durand and Bill Freeman ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400014290 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111213T130000 DTEND;TZID="Asia/Jerusalem":20111213T150000 DTSTAMP;TZID="Asia/Jerusalem":20111213T130000 FREEBUSY;FBTYPE=BUSY:20111213T130000/20111213T150000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Jasmin Fisher (Mic rosoft Research, Cambridge, UK) about Bioinformatics Forum: POSTPONED! at 2011-12-13 13:00:00 DESCRIPTION;LANGUAGE=en-US:The decade of genomic revolution following the human genome's sequencing has produced significant medical advances, and yet again, revealed how complicated human biology is, and how much more remains to be understood. Biology is an extraordinary complicated puzzl e; we may know some of its pieces but have no clue how they are assemble d to orchestrate the symphony of life, which renders the comprehension a nd analysis of living systems a major challenge. Recent efforts to creat e executable models of complex biological phenomena - an approach we cal l Executable Biology - entail great promise for new scientific discoveri es, shading new light on the puzzle of life. At the same time, this new wave of the future forces computer science to stretch far and beyond, an d in ways never considered before, in order to deal with the enormous co mplexity observed in biology. This talk will focus on our recent success stories in using formal methods to model cell fate decisions during dev elopment and cancer, and ongoing efforts to develop dedicated tools for biologists to model cellular processes in a visual friendly way. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400014190 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111214T113000 DTEND;TZID="Asia/Jerusalem":20111214T133000 DTSTAMP;TZID="Asia/Jerusalem":20111214T113000 FREEBUSY;FBTYPE=BUSY:20111214T113000/20111214T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Shlomi Dolev (Math and CS, Ben G urion University) about ceClub: Multi-party Computation Forever, for Cloud Computing and Beyond at 2011-12-14 11:30:00 DESCRIPTION;LANGUAGE=en-US:Three works will be described. In the first we present reactive secret sharing, that changes the secret according to unbo unded sequence of common inputs, where no communication among the (dynamic set of) participants is allowed, we present a fully secure solution for s imple functions but somewhat non secure solution for any function.. In the second work dynamic on-going multiparty computation, in which we consider the case of dynamic group of participants that should not know the sequen ce of inputs of the others, as well as should not know the function perfor med on the inputs. In the last work we consider the case of infinite execu tion with no communication among the participants where we prove that any automaton can be executed in a fully secure fashion, the construction is b ased on Krohn-Rhodes decomposition technique.

Joint works wi th Limor Lahiani, Moti Yung, Juan Garay, Niv Gilboa and Vladimir Kolesniko v

bio: Shlomi Dolev received his B.Sc. in Engineering and B. A. in Computer Science in 1984 and 1985, and his M.Sc. and D.Sc. in comput er Science in 1990 and 1992 from the Technion Israel Institute of Technolo gy. From 1992 to 1995 he was at Texas A&M University postdoc of Jennifer W elch. In 1995 he joined the Department of Mathematics and Computer Science at Ben-Gurion University where he is now a full professor and the dean of natural sciences.

He was a visiting researcher/professor at MIT, DIMACS, and LRI, for several periods during summers. Shlomi is the a uthor of the book "self-stabilization" published by the MIT Press. He publ ished two hundreds journal and conference scientific articles, and patents . Shlomi served in the program committee of more than 60 conferences inclu ding: the ACM Symposium on Principles of Distributed Computing, and the In ternational Symposium on DIStributed Computing. He is an associate editor of the IEEE Transactions on Computers, the AIAA Journal of Aerospace Compu ting, Information and Communication and a guest editor of the Distributed Computing Journal and the Theoretical Computer Science Journal. His resear ch grants include IBM faculty awards, Intel academic grants, Verisign, ISF , US Airforce, EU and NSF grants.

Shlomi is the founding cha ir of the computer science department at Ben-Gurion University, where he n ow holds the Rita Altura trust chair in computer science. His current rese arch interests include distributed computing, distributed systems, securit y and cryptography and communication networks; in particular the self-stab ilization property of such systems. Recently, he is involved in optical co mputing research. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400014310 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111214T123000 DTEND;TZID="Asia/Jerusalem":20111214T143000 DTSTAMP;TZID="Asia/Jerusalem":20111214T123000 FREEBUSY;FBTYPE=BUSY:20111214T123000/20111214T143000 SUMMARY;LANGUAGE=en-US:The msc talk by Ariel Raviv about Concept-based App roach to Word-Sense Disambiguation at 2011-12-14 12:30:00 DESCRIPTION;LANGUAGE=en-US:The task of automatically determining the corre ct sense of a polysemous word has remained a challenge to this day. It is crucial in many Natural Language Processing (NLP) applications such as spe ech recognition, information retrieval, machine translation and computatio nal advertising. In our research, we introduce Concept-Based Disambiguatio n (CBD), a novel framework that utilizes recent semantic analysis techniqu es to represent both the context of the word and its senses in a high-dime nsional space of natural concepts. The concepts are retrieved from a vast encyclopedic resource, thus enriching the disambiguation process with larg e amount of domain-specific knowledge. In such concept-based spaces, more comprehensive measures can be applied in order to pick the right sense. Ad ditionally, we introduce a novel representation scheme, denoted Anchored R epresentation, that builds a more specific representation for a text that is associated with an anchoring word. We evaluated our framework using two recent text representation schemes, Explicit Semantic Analysis (ESA) and Compact Hierarchical Explicit Semantic Analysis (CHESA) and their two anch ored counterparts, and showed that the anchored representation is more sui table to the task of Word sense Disambiguation (WSD). Finally we evaluate our system in coarse-grained settings and show that it outperforms state-o f-the-art learning-based methods that exploit large annotated corpora and that it is comparable to recent state-of-the-art unsupervised methods that make use of vast knowledge bases. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400014280 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111214T123000 DTEND;TZID="Asia/Jerusalem":20111214T143000 DTSTAMP;TZID="Asia/Jerusalem":20111214T123000 FREEBUSY;FBTYPE=BUSY:20111214T123000/20111214T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Gil Cohen (Weizmann Inst itute for Science) about Theory Seminar: Non-Malleable Extractors with Sho rt Seeds and Applications to Privacy Amplification at 2011-12-14 12:30: 00 DESCRIPTION;LANGUAGE=en-US:Motivated by the classical problem of privacy a mplification, Dodis and Wichs (STOC '09) introduced the notion of a non- malleable extractor, significantly strengthening the notion of a strong extractor. A non-malleable extractor is a function $\nmExt : \{0,1\}^n \ times \{0,1\}^d \rightarrow \{0,1\}^m$ that takes two inputs: a weak sou rce $W$ and a uniform (independent) seed $S$, and outputs a string $\nmE xt(W,S)$ that is nearly uniform given the seed $S$ as well as the value $\nmExt(W, S')$ for any seed $S' \neq S$ that may be determined as an ar bitrary function of $S$.

The first explicit construction of a non-malleable extractor was recently provided by Dodis, Li, Wooley an d Zuckerman (FOCS '11). Their extractor works for any weak source with m in-entropy rate $1/2 + \delta$, where $\delta > 0$ is an arbitrary cons tant, and outputs up to a linear number of bits, but suffers from two dr awbacks. First, the length of its seed is linear in the length of the we ak source (which leads to privacy amplification protocols with high comm unication complexity). Second, the construction is conditional: when out putting more than a logarithmic number of bits (as required for privacy amplification protocols) its efficiency relies on a longstanding conject ure on the distribution of prime numbers.

In this work we p resent an unconditional construction of a non-malleable extractor with s hort seeds. For any integers $n$ and $d$ such that $2.01 \cdot \log{n} \ le d \le n$, we present an explicit construction of a non-malleable extr actor $\nmExt \colon \{0,1\}^n \times \{0,1\}^d \rightarrow \{0,1\}^m$, with $m=\Omega(d)$, and error exponentially small in $m$. The extractor works for any weak source with min-entropy rate $1/2 + \delta$, where $\ delta > 0$ is an arbitrary constant. Moreover, our extractor in fact sa tisfies an even more general notion of non-malleability: its output $\nm Ext(W,S)$ is nearly uniform given the seed $S$ as well as the values $\n mExt(W, S_1), \ldots, \nmExt(W, S_t)$ for several seeds $S_1, \ldots, S_ t$ that may be determined as an arbitrary function of $S$, as long as $S \notin \{S_1, \ldots, S_t\}$. By instantiating the framework of Dodis a nd Wichs with our non-malleable extractor, we obtain the first $2$-round privacy amplification protocol for min-entropy rate $1/2 + \delta$ with asymptotically optimal entropy loss and poly-logarithmic communication complexity. This improves the previously known $2$-round privacy amplifi cation protocols: the protocol of Dodis and Wichs whose entropy loss is not asymptotically optimal, and the protocol of Dodis, Li, Wooley and Zu ckerman whose communication complexity is linear.

Joint wor k with Ran Raz and Gil Segev. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400014320 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111218T113000 DTEND;TZID="Asia/Jerusalem":20111218T133000 DTSTAMP;TZID="Asia/Jerusalem":20111218T113000 FREEBUSY;FBTYPE=BUSY:20111218T113000/20111218T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Eldad Haber (Math & Earth an d Ocean Science, University of British Columbia) about Pixel Club: An Eff ective Method for Parameter Estimation with PDE Constraints with Multiple Right Hand Sides at 2011-12-18 11:30:00 DESCRIPTION;LANGUAGE=en-US:Many parameter estimation problems involve a pa rameter dependent PDEs with multiple right hand sides. The computational cost and memory requirements of such problems increase linearly with th e number of right hand sides. For many applications this is the main bot tleneck of the computation. In this talk we show that problems with mult iple right hand sides can be reformulated as stochastic optimization pro blems that are much cheaper to solve. We discuss the solution methodolog y and use the direct current resistivity and seismic tomography as model problems to show the effectiveness of our approach. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400014370 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111218T143000 DTEND;TZID="Asia/Jerusalem":20111218T163000 DTSTAMP;TZID="Asia/Jerusalem":20111218T143000 FREEBUSY;FBTYPE=BUSY:20111218T143000/20111218T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Aviv Zohar about Challenges in M ulti-Agent Systems: Bitcoin, Social Networks, P2P Communities, and Network Protocols. at 2011-12-18 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014200 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111220T123000 DTEND;TZID="Asia/Jerusalem":20111220T143000 DTSTAMP;TZID="Asia/Jerusalem":20111220T123000 FREEBUSY;FBTYPE=BUSY:20111220T123000/20111220T143000 SUMMARY;LANGUAGE=en-US:The CSpecial Talk talk by Alon Shalita (Facebook) a bout CSpecial Talk: Innovation and Scale at Facebook at 2011-12-20 12:30:0 0 DESCRIPTION;LANGUAGE=en-US:In seven years of rapid growth Facebook has bec ome today's most popular social network, redefining digital identity an d the basic set of communication channels for people worldwide. Many un ique engineering challenges were handled along the way: building produc ts extremely quickly while keeping the high quality of the service and the low cost of the hardware that supports it, allowing hori-zontal sca ling of highly interconnected datar ,anking news items for divergent us er population with and allowing application )from a social revolution t o the party next door( large spectrum of interestsdevelopers and websit e owners to extend the social graph using an open platform API.

In this talk, we will peek inside Facebook engineering, describing some of these challenges and their solutions. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014440 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111220T133000 DTEND;TZID="Asia/Jerusalem":20111220T153000 DTSTAMP;TZID="Asia/Jerusalem":20111220T133000 FREEBUSY;FBTYPE=BUSY:20111220T133000/20111220T153000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Jasmin Fisher (Mic rosoft Research, Cambridge, UK) about Bioinformatics Forum: From Coding th e Genome to Algorithms Decoding Life at 2011-12-20 13:30:00 DESCRIPTION;LANGUAGE=en-US:The decade of genomic revolution following the human genome's sequencing has produced significant medical advances, and yet again, revealed how complicated human biology is, and how much more remains to be understood. Biology is an extraordinary complicated puzzl e; we may know some of its pieces but have no clue how they are assemble d to orchestrate the symphony of life, which renders the comprehension a nd analysis of living systems a major challenge. Recent efforts to creat e executable models of complex biological phenomena - an approach we cal l Executable Biology - entail great promise for new scientific discoveri es, shading new light on the puzzle of life. At the same time, this new wave of the future forces computer science to stretch far and beyond, an d in ways never considered before, in order to deal with the enormous co mplexity observed in biology. This talk will focus on our recent success stories in using formal methods to model cell fate decisions during dev elopment and cancer, and ongoing efforts to develop dedicated tools for biologists to model cellular processes in a visual friendly way. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400014360 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111220T143000 DTEND;TZID="Asia/Jerusalem":20111220T163000 DTSTAMP;TZID="Asia/Jerusalem":20111220T143000 FREEBUSY;FBTYPE=BUSY:20111220T143000/20111220T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Ohad Shamir about Machine Learni ng: Higher, Faster, Stronger at 2011-12-20 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014150 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111220T143000 DTEND;TZID="Asia/Jerusalem":20111220T163000 DTSTAMP;TZID="Asia/Jerusalem":20111220T143000 FREEBUSY;FBTYPE=BUSY:20111220T143000/20111220T163000 SUMMARY;LANGUAGE=en-US:The CSpecial Talk talk by Ariel Kogan (BMC Software ) about CSpecial Talk: Building Software You Can Trust at 2011-12-20 14:30 :00 DESCRIPTION;LANGUAGE=en-US:We all write code, as a matter of fact, for mos t of us is part of what we do on a daily basis be it our jobs, personal pr ojects or studies. However, when all those lines of code that we throw her e and there become part of an application and that application is develope d by a significant number of developers, we should take proper care of our code.

Many names have been given to proper management of cod e: Configuration Manager, Application Lifecycle Manager, DevOps, etc. and many tools have been developed to make our life easier. The truth is that no matter if someone is doing it for you in your organization, every devel oper should be aware of it and fully understands the process.

Why do we need to do this? Because we are probably going to be maintainin g more than one version of our product at any given time and we want to be able to trace back everything we did to help us understand it or fix it. We most likely may also be required to be able to rollback certain parts o f our code fast and easy for numerous different reasons.

In t his talk we are going to go through the entire stack that will let us mana ge in a relatively simple way our code/application lifecycle. We are going to start from the IDE, we’ll keep track of our code, build it, manage d ependencies, test it, produce reports, deploy it and handle bugs or change requests, everything will be done in a traceable manner. It’s an attrac tive session that will give us an overview of the practices that are a mus t in every (serious) company you’ll work. In addition to the formal pres entation we will have a live ongoing demo for a hands-on experience.

Some of the tools we are going to introduce are: Git, Gradle, Ecli pse, IntelliJ IDEA, Jenkins, Artifactory, JUnit, Cobertura and Jira. Most of them are free to use and strongly backed by the Community.

Join us for the session that will change the way you develop. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 7 UID:111ec4410400014330 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111221T113000 DTEND;TZID="Asia/Jerusalem":20111221T133000 DTSTAMP;TZID="Asia/Jerusalem":20111221T113000 FREEBUSY;FBTYPE=BUSY:20111221T113000/20111221T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Asaf Cidon (EE, Stanford Univers ity) about ceClub: Flashback: A New Control Channel for Wireless Networks at 2011-12-21 11:30:00 DESCRIPTION;LANGUAGE=en-US:Unlike wired network protocols, Wi-Fi does not separate the data channel from the control channel, since only a single se nder and receiver can communicate at a given time slot. Flashback is a sys tem that allows multiple transmitters to send 'flashes' of high power OFDM sub-carriers without affecting the normal data transmissions on the Wi-Fi channel. By taking advantage of SNR margins of Wi-Fi channel codes, the f lashes do not impose any overhead on the regular data transmission. Using this simple PHY mechanism, Flashback provides higher networking layers wit h a multi-user control channel, which can be used to significantly improve existing Wi-Fi protocols. During the talk we will share the results from our implementation of Flashback, and discuss their implications on the hig her layers of the Wi-Fi stack.

bio: Asaf Cidon is an Electri cal Engineering PhD and MBA student at Stanford University, where he condu cts research on data-center and mobile systems under Professors Mendel Ros enblum and Sachin Katti. He worked at Google Israel and helped develop Goo gle Related and previously served as a Product Manager for two start-ups. Prior to his studies, Asaf served as a team leader in an elite unit in the Israeli Intelligence Forces. He received his BSc in Computer and Software Engineering Cum Laude from the Technion, and is a recipient of the Stanfo rd Graduate Fellowship and Sohnis Promising Scientist Award. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 861 UID:111ec4410400014380 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111221T113000 DTEND;TZID="Asia/Jerusalem":20111221T133000 DTSTAMP;TZID="Asia/Jerusalem":20111221T113000 FREEBUSY;FBTYPE=BUSY:20111221T113000/20111221T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Tomer London (EE, Stanford Unive rsity) about ceClub: MARS: Adaptive Remote Execution Scheduler for Multith readed Mobile Devices at 2011-12-21 11:30:00 DESCRIPTION;LANGUAGE=en-US:Mobile devices face a growing demand to support computationally intensive applications like 3D graphics and computer visi on. However, these devices are inherently limited by processor power densi ty and device battery life. Dynamic remote execution addresses this proble m, by enabling mobile devices to opportunistically offload computations to a remote server. We envision remote execution as a new type of cloud-base d heterogeneous computing resource, or a "Cloudon-Chip", which would be ma naged as a system resource as if it were a local CPU, with a highly variab le wireless interconnect. To realize this vision, we introduce MARS, the f irst adaptive, online and lightweight RPC-based remote execution scheduler supporting multi-threaded and multi-core systems. MARS uses a novel effic ient offloading decision algorithm that takes into account the inherent tr ade-offs between communication and computation delays and power consumptio n. Due to its lightweight design, MARS runs on the device itself, instantl y adapts its decisions to changing wireless resources, and supports any nu mber of threads and cores. We evaluated MARS using a trace-based simulator driven by real world measurements on augmented reality, face recognition and video game applications. MARS achieves an average speedup of 57% and 3 3% higher energy savings over the best static client-server partitions. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 861 UID:111ec4410400014420 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111221T123000 DTEND;TZID="Asia/Jerusalem":20111221T143000 DTSTAMP;TZID="Asia/Jerusalem":20111221T123000 FREEBUSY;FBTYPE=BUSY:20111221T123000/20111221T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Danny Hermelin (Max-Plan ck Institut Informatik, Saarbrucken, Germany) about Theory Seminar: (In) C ompressibility of NP-hard Problems at 2011-12-21 12:30:00 DESCRIPTION;LANGUAGE=en-US:A compression algorithm for a computation probl em is a polynomial-time algorithm that compresses instances of the given problem into equivalent instances. The performance of the compression i s naturally measured with respect to its worst-case output size. While N P-hard problems cannot have compression algorithms with non-trivial perf ormance guarantees in terms of the original input size (assuming NP is n ot in P), some NP-hard problems have surprisingly good compressions when performance is measured in terms of the input solution size. In this ta lk we discuss some recently developed lower bounds for the compressibili ty of NP-hard problems when the latter measure is used. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400014390 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111221T143000 DTEND;TZID="Asia/Jerusalem":20111221T163000 DTSTAMP;TZID="Asia/Jerusalem":20111221T143000 FREEBUSY;FBTYPE=BUSY:20111221T143000/20111221T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Dan Roth SPECIAL LECTURE, Note u nusual day about Learning from Natural Instructions at 2011-12-21 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014060 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111222T103000 DTEND;TZID="Asia/Jerusalem":20111222T123000 DTSTAMP;TZID="Asia/Jerusalem":20111222T103000 FREEBUSY;FBTYPE=BUSY:20111222T103000/20111222T123000 SUMMARY;LANGUAGE=en-US:The colloq talk by Guy Rothblum about How to Comput e in the Presence of Leakage at 2011-12-22 10:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. (NOTE UNUSUAL TIME! UID:111ec4410400014400 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111222T143000 DTEND;TZID="Asia/Jerusalem":20111222T163000 DTSTAMP;TZID="Asia/Jerusalem":20111222T143000 FREEBUSY;FBTYPE=BUSY:20111222T143000/20111222T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Eden Chlamtac about Convex Progr amming Hierarchies: Trading Time for Approximation at 2011-12-22 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014410 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111225T143000 DTEND;TZID="Asia/Jerusalem":20111225T163000 DTSTAMP;TZID="Asia/Jerusalem":20111225T143000 FREEBUSY;FBTYPE=BUSY:20111225T143000/20111225T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Lee-Ad Gottlieb about The travel ing salesman problem: Low-dimensionality implies a polynomial time approxi mation scheme. at 2011-12-25 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014260 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111226T183000 DTEND;TZID="Asia/Jerusalem":20111226T203000 DTSTAMP;TZID="Asia/Jerusalem":20111226T183000 FREEBUSY;FBTYPE=BUSY:20111226T183000/20111226T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Ofer Rosenberg (AMD ) about Haifux Club: GPGPU - Case studies, Do's and Dont's (Part 3 out of 4 talks series) at 2011-12-26 18:30:00 DESCRIPTION;LANGUAGE=en-US:This is a 4 series of 4 talks about GPGPUS, int ended for the practical engineer:
1. Motivation, AMD's architecture
2. OpenCL
3.Case studies, Dos and Don'ts
4.Tools and Profiling for Performance

General Purpose GPU programming be came a hot topic in the last few years, ranging from academic studies to b eing used by commercial software products. As an example, three out of the world's top10 supercomputers (June2011 list) contain GPUs in them. This s eries of lectures focuses on OpenCL, the open standard for parallel progra mming of heterogeneous systems. The third and fourth meetings will incl ude hands-on experience, under Ofer's guidance. For these meetings, each p articipant will have to bring a laptop with:
1. Linux (this is Haif ux, the Haifa Linux Club)
2. An x86 processor, either AMD or Intel. An IPAD is not good for this purpose.
3. AMD's OpenCL implementatio n installed. It works both on AMD and Intel, installation details for thos e who need them will follow.
4. Eclipse installed.
5. If poss ible, an AMD GPU, with a version number of 54 or higher. Use "lspci | grep VGA" to find the GPU vendor & model.

To arrange for machine acc ess for those who do not have a suitable laptop, as well as prepare for th e series, please email in private to webmaster@haifux.org
• If you intend to attend the seri es
• If you have (or do not have) the required gear. In particular , if you have an AMD GPU.

Syllabus: Introduction to GPGPU Programming

General Purpose GPU programming became a hot t opic in the last few years, ranging from academic studies to being used by commercial software products. As an example, three out of the world 's top10 supercomputers (June2011 list) contain GPUs in them. This seri es of lectures focuses on OpenCL, the open standard for parallel progra mming of heterogeneous systems.


1. GPGPU introduction

The first lecture is an introduction to GPU architecture and GPGPU progra ming. It covers the differences between GPU and CPU architectures, and how these differences impose restrictions on programming GPUs. We will also touch the issue of memory aspects of GPU architecture and the over all system (CPU & GPU)


2. OpenCL overview

From the Khr onos website: "OpenCL" is the first open, royalty-free standard for cro ss-platform, parallel programming of modern processors found in personal computers, servers and handheld/embedded devices. This lecture will prov ide an overview of OpenCL, covering the API programming aspects (such a s OpenCL objects, contexts, queues, events, etc.) as well as the langua ge enhancements (such as vectors, images, samplers, built-in functions etc.)


3. OpenCL Do's and Don'ts

This lecture provides a practical guide for programming in OpenCL by doing a hands-on guided experience of writing OpenCL applications and kernels. Starting from ba sic examples through more complex scenarios, we will provide some tips for writing code that provides the required correct results. We will also provide some performance tips.


4. OpenCL Optimization & Pro filing

This lecture focuses on performance aspects of OpenCL. We w ill provide a hands-on experience of improving performance of OpenCL ke rnels by optimizing a specific example. In addition we will show ways t o profile the kernel, including working with profiling tools such as AM D kernel profiler and gDebugger. Note that some of issues presented in this lecture will be possible only on AMD GPUs.

Lecture slides in PDF format

Back to the Club's homepage ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400014220 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111227T103000 DTEND;TZID="Asia/Jerusalem":20111227T123000 DTSTAMP;TZID="Asia/Jerusalem":20111227T103000 FREEBUSY;FBTYPE=BUSY:20111227T103000/20111227T123000 SUMMARY;LANGUAGE=en-US:The CSpecial Talk talk by Yechiel Kimchi about CSp ecial Talk: Coding Standards for Software Correctness at 2011-12-27 10:30: 00 DESCRIPTION;LANGUAGE=en-US:

From an Engineering stand point it 's difficult to argue against the view that software tools should meet, fi rst and foremost, their specifications. This talk concentrates on remindi ng us a fundamental technique for achieving this goal, while arguing that some basic programming idioms, which most are well known (but not as well followed), can make this technique easier to handle.

This par t is an expansion of the terse version I've contributed to the book " 97 Things Every Programmer Should Know" (O'Reilly, 2010) http://programmer.97things.oreilly.com/wiki/index.php/Coding_with_Rea son

Along the way I'll mention some of the more common (a nd very useful) techniques that are used for managing the quality of indus trial software and, when relevant, will mention a few shortcomings of them .

Parts of the contents of the presentation have appeared as an article on C-Vu, a journal of ACCU (the Association of C and C++ Users) , pp. 8-11, May 2011, by the same title. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400014500 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111227T143000 DTEND;TZID="Asia/Jerusalem":20111227T163000 DTSTAMP;TZID="Asia/Jerusalem":20111227T143000 FREEBUSY;FBTYPE=BUSY:20111227T143000/20111227T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Ilan Gronau about Using Individu al Human Genomes to Illuminate the Mysteries of Early Human History at 201 1-12-27 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014300 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111228T113000 DTEND;TZID="Asia/Jerusalem":20111228T133000 DTSTAMP;TZID="Asia/Jerusalem":20111228T113000 FREEBUSY;FBTYPE=BUSY:20111228T113000/20111228T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Yuval Emek (ETH) about Pixel Clu b: Distributed Computing: Bridging the Gap Between Theory and Practice at 2011-12-28 11:30:00 DESCRIPTION;LANGUAGE=en-US:The theory of distributed computing, which lies at the heart of understanding the power and limitations of distributed sy stems, underwent tremendous progress over the last few decades. Despite th is progress, there seems to be a widening gap between the traditional mode ls on top of which the theory of distributed computing is built and the re al-world problems we wish to investigate through these models. In this tal k we will discuss the different aspects of this widening gap and present s ome of the efforts made in attempt to adjust the field of theoretical dist ributed computing to the rapidly changing needs of its practical applicati ons. In particular, we will focus on recent advances in the study of wirel ess networks and on a new model for decentralized networks of "unorthodox" devices. The talk will be self-contained. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 861 UID:111ec4410400014470 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111228T123000 DTEND;TZID="Asia/Jerusalem":20111228T143000 DTSTAMP;TZID="Asia/Jerusalem":20111228T123000 FREEBUSY;FBTYPE=BUSY:20111228T123000/20111228T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Dvir Falik (The Hebrew U niversity of Jerusalem) about Theory Seminar: An algebraic Proof of a Robu st Social Choice Impossibility Theorem at 2011-12-28 12:30:00 DESCRIPTION;LANGUAGE=en-US:An important element of social choice theory ar e impossibility theorems, such as Arrow's theorem and Gibbard-Satterthwa ite's theorem, which state that under certain natural constraints, socia l choice mechanisms are impossible to construct. In recent years, much w ork has been done in finding robust versions of these theorems, showing that impossibility remains even when the constraints are almost always satisfied. In this work we present a general spectral technique for tac kling such problems, and demonstrate it on a variant of Arrow's theorem. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400014480 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111229T093000 DTEND;TZID="Asia/Jerusalem":20111229T113000 DTSTAMP;TZID="Asia/Jerusalem":20111229T093000 FREEBUSY;FBTYPE=BUSY:20111229T093000/20111229T113000 SUMMARY;LANGUAGE=en-US:The CSpecial Talk talk by Nir Piterman (University of Leicester) about CSpecial Talk: Synthesis From Temporal Specifications at 2011-12-29 09:30:00 DESCRIPTION;LANGUAGE=en-US:In this talk I give a short introduction to the process of synthesis, the automatic production of designs from their sp ecifications. We are interested in reactive systems, systems that contin uously interact with other programs, users, or their environment and s pecifications in linear temporal logic. Classical solutions to synthesis use either two player games or tree automata. We give a short introdu ction to the technique of using two player games for synthesis and how t o solve such games.

The classical solution to synthesis requi res the usage of deterministic automata. This solution is 2EXPTIME-compl ete, is quite complicated, and does not work well in practice. We sugg est a syntactic approach that restricts the kind of properties users are allowed to write.

We claim that this approach is general enough and can be extended to cover most properties written in practice. The main advantage of our approach is that it is tailored to the use of BDD s and uses the structure of given properties to handle them more efficie ntly. We discuss how to extend our approach to handle more general pr operties and mention future directions. Finally, we give a few example s of applications from hardware design, model-driven development, and ro bot control. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014430 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111229T113000 DTEND;TZID="Asia/Jerusalem":20111229T133000 DTSTAMP;TZID="Asia/Jerusalem":20111229T113000 FREEBUSY;FBTYPE=BUSY:20111229T113000/20111229T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Eitan Sharon (Videosurf) abo ut Pixel Club: VideoSurf's video Recognition Technology in the Connected Devices Race at 2011-12-29 11:30:00 DESCRIPTION;LANGUAGE=en-US:Online video content has grown enormously, such that it is a huge proportion of newly-created content. There are an est imated four billion different video links available to consumers for wat ching online. YouTube users alone are currently uploading about half a m illion videos a day. The leading mobile platforms iOS, Android and Windo ws are provided with ever larger phone screens for convenient video view ing, and even more so are the various tablets such as the iPad. Furtherm ore, the industry has only just begun the competition over the online TV experience (e.g. Google TV, Xbox Live), together with all its companion apps (e.g. Xfinity).

In this new era of video-content explosion the need to organize all of the world's videos for optimal discovery an d viewing dictates that cutting-edge visual recognition technologies be applied. For scalability, it is impractical to spend more than a couple of milliseconds processing a single video frame. Videosurf (of which I a m chief technology officer) is a silicon valley company recently sold to Microsoft that has tackled some of these issues. In this talk I will de scribe a few of the video analysis techniques that Videosurf has develop ed to facilitate discovery and viewing of online video content. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400014460 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20111229T143000 DTEND;TZID="Asia/Jerusalem":20111229T163000 DTSTAMP;TZID="Asia/Jerusalem":20111229T143000 FREEBUSY;FBTYPE=BUSY:20111229T143000/20111229T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Avi Cohen about Chamsa" and Comp uter Science at 2011-12-29 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014350 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120101T143000 DTEND;TZID="Asia/Jerusalem":20120101T163000 DTSTAMP;TZID="Asia/Jerusalem":20120101T143000 FREEBUSY;FBTYPE=BUSY:20120101T143000/20120101T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Shahar Maoz about Semantic Model Differencing for Software Evolution at 2012-01-01 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014340 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120102T183000 DTEND;TZID="Asia/Jerusalem":20120102T203000 DTSTAMP;TZID="Asia/Jerusalem":20120102T183000 FREEBUSY;FBTYPE=BUSY:20120102T183000/20120102T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Ofer Rosenberg and Yaki Tebeka (AMD) about Haifux Club: GPGPU - Case studies, Do's and Dont' s (Part 4 out of 4 talks series) at 2012-01-02 18:30:00 DESCRIPTION;LANGUAGE=en-US:This is a 4 series of 4 talks about GPGPUS, int ended for the practical engineer:
1. Motivation, AMD's architecture
2. OpenCL
3.Case studies, Dos and Don'ts
4.Tools and Profiling for Performance

General Purpose GPU programming be came a hot topic in the last few years, ranging from academic studies to b eing used by commercial software products. As an example, three out of the world's top10 supercomputers (June2011 list) contain GPUs in them. This s eries of lectures focuses on OpenCL, the open standard for parallel progra mming of heterogeneous systems. The third and fourth meetings will incl ude hands-on experience, under Ofer's guidance. For these meetings, each p articipant will have to bring a laptop with:
1. Linux (this is Haif ux, the Haifa Linux Club)
2. An x86 processor, either AMD or Intel. An IPAD is not good for this purpose.
3. AMD's OpenCL implementatio n installed. It works both on AMD and Intel, installation details for thos e who need them will follow.
4. Eclipse installed.
5. If poss ible, an AMD GPU, with a version number of 54 or higher. Use "lspci | grep VGA" to find the GPU vendor & model.

To arrange for machine acc ess for those who do not have a suitable laptop, as well as prepare for th e series, please email in private to webmaster@haifux.org
• If you intend to attend the seri es
• If you have (or do not have) the required gear. In particular , if you have an AMD GPU.

Syllabus: Introduction to GPGPU Programming

General Purpose GPU programming became a hot t opic in the last few years, ranging from academic studies to being used by commercial software products. As an example, three out of the world 's top10 supercomputers (June2011 list) contain GPUs in them. This seri es of lectures focuses on OpenCL, the open standard for parallel progra mming of heterogeneous systems.


1. GPGPU introduction

The first lecture is an introduction to GPU architecture and GPGPU progra ming. It covers the differences between GPU and CPU architectures, and how these differences impose restrictions on programming GPUs. We will also touch the issue of memory aspects of GPU architecture and the over all system (CPU & GPU)


2. OpenCL overview

From the Khr onos website: "OpenCL" is the first open, royalty-free standard for cro ss-platform, parallel programming of modern processors found in personal computers, servers and handheld/embedded devices. This lecture will prov ide an overview of OpenCL, covering the API programming aspects (such a s OpenCL objects, contexts, queues, events, etc.) as well as the langua ge enhancements (such as vectors, images, samplers, built-in functions etc.)


3. OpenCL Do's and Don'ts

This lecture provides a practical guide for programming in OpenCL by doing a hands-on guided experience of writing OpenCL applications and kernels. Starting from ba sic examples through more complex scenarios, we will provide some tips for writing code that provides the required correct results. We will also provide some performance tips.


4. OpenCL Optimization & Pro filing

This lecture focuses on performance aspects of OpenCL. We w ill provide a hands-on experience of improving performance of OpenCL ke rnels by optimizing a specific example. In addition we will show ways t o profile the kernel, including working with profiling tools such as AM D kernel profiler and gDebugger. Note that some of issues presented in this lecture will be possible only on AMD GPUs.

Lecture slides in PDF format

Back to the Club's homepage ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400014450 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120103T113000 DTEND;TZID="Asia/Jerusalem":20120103T133000 DTSTAMP;TZID="Asia/Jerusalem":20120103T113000 FREEBUSY;FBTYPE=BUSY:20120103T113000/20120103T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Daniel Glasner (Math & CS, T he Weizmann Institute of Science) about Pixel Club: Contour-Based Joint C lustering of Multiple Segmentations at 2012-01-03 11:30:00 DESCRIPTION;LANGUAGE=en-US:We present an unsupervised, shape-based method for joint clustering of multiple image segmentations. Given two or more closely-related images, such as close frames in a video sequence or imag es of the same scene taken under different lighting conditions, our meth od generates a joint segmentation of the images. We introduce a novel co ntour-based representation that allows us to cast the shape-based joint clustering problem as a quadratic semi-assignment problem. Our score fun ction is additive. We use complex-valued affinities to assess the qualit y of matching the edge elements at the exterior bounding contour of clus ters, while ignoring the contributions of elements that fall in the inte rior of the clusters. We further combine this contour-based score with r egion information and use a linear programming relaxation to solve for the joint clusters. We evaluate our approach on the occlusion boundary data -set of Stein et al.

This is joint work with Shiv N. Vitala devuni and Ronen Basri. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400014530 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120103T124500 DTEND;TZID="Asia/Jerusalem":20120103T144500 DTSTAMP;TZID="Asia/Jerusalem":20120103T124500 FREEBUSY;FBTYPE=BUSY:20120103T124500/20120103T144500 SUMMARY;LANGUAGE=en-US:The ceClub talk by Erich Nahum (IBM T.J. Watson Res earch Center) about ceClub: Load Balancing for SIP Server Clusters at 2012 -01-03 12:45:00 DESCRIPTION;LANGUAGE=en-US:The Session Initiation Protocol (SIP) is widely used for controlling communication sessions such as voice and video calls over IP, video conferencing, streaming multimedia distribution, instant m essaging, presence information, file transfer and online games. This work introduces several novel load balancing algorithms for distributing SIP re quests to a cluster of SIP servers. Our load balancer improves both throug hput and response time versus a single node, while exposing a single inter face to external clients. We present the design, implementation and evalua tion of our system using a cluster of Intel x86 machines running Linux. We compare our algorithms with several well-known approaches and present sca lability results for up to 10 nodes. Our best algorithm, Transaction Least -Work-Left (TLWL), achieves its performance by integrating several feature s: knowledge of the SIP protocol; dynamic estimates of back-end server loa d; distinguishing transactions from calls; recognizing variability in call length; and exploiting differences in processing costs for different SIP transactions. By combining these features, our algorithm provides finer-gr ained load balancing than standard approaches, resulting in throughput imp rovements of up to 24 percent and response time improvements of up to two orders of magnitude. We present a detailed analysis of occupancy to show h ow our algorithms significantly reduce response time.

Referen ce: This work appeared in IEEE INFOCOM in 2009 and is accepted to appear i n IEEE/ACM Transactions on Networking. This research is joint with Hongbo Jiang of Huazong University of Science and Technology, and Arun Iyengar, W olfgang Segmuller, Asser Tantawi, and Charles P. Wright of the IBM T.J. Wa tson Research Center. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400014560 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120103T143000 DTEND;TZID="Asia/Jerusalem":20120103T163000 DTSTAMP;TZID="Asia/Jerusalem":20120103T143000 FREEBUSY;FBTYPE=BUSY:20120103T143000/20120103T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Andrei Bulatov about The complex ity of counting constraint satisfaction problem at 2012-01-03 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014490 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120104T113000 DTEND;TZID="Asia/Jerusalem":20120104T133000 DTSTAMP;TZID="Asia/Jerusalem":20120104T113000 FREEBUSY;FBTYPE=BUSY:20120104T113000/20120104T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Rob Bisseling (Utrecht Universit y) about ceClub: Graph Matching and Clustering on the GPU at 2012-01-04 11 :30:00 DESCRIPTION;LANGUAGE=en-US:Graph matching is the problem of matching nodes of a graph in pairs such that the largest number of pairs is created or t he largest sum of edge weights is obtained. Greedy graph matching provides us with a fast way to coarsen a given graph during graph partitioning. Di rect algorithms on the CPU which perform such greedy matchings are simple and fast, but offer few handholds for parallelisation. To remedy this, we introduce a fine-grained shared-memory parallel algorithm for greedy match ing, together with an implementation on the GPU, which is faster than the CPU algorithms and produces matchings of similar or better quality.

Another graph problem with much practical interest is that of grap h clustering, where highly connected clusters with many internal edges and few external edges are determined. Agglomerative clustering is an effecti ve greedy way to quickly generate graph clusterings of high quality. We in troduce a fine-grained shared memory algorithm for agglomerative clusterin g on both the CPU and GPU. This heuristic algorithm is able to generate cl usterings in very little time: an optimal clustering is obtained from a st reet network graph with 14 million vertices and 17 million edges in six se conds on the GPU. The clustering work was done as part of the 10th DIMACS challenge on graph partitioning and clustering.

Joint work w ith Bas Fagginger Auer. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 9 UID:111ec4410400014540 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120104T123000 DTEND;TZID="Asia/Jerusalem":20120104T143000 DTSTAMP;TZID="Asia/Jerusalem":20120104T123000 FREEBUSY;FBTYPE=BUSY:20120104T123000/20120104T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Ron Rothblum (Weizmann I nstitute for Science) about Theory Seminar: Homomorphic Encryption: from P rivate-Key to Public-Key at 2012-01-04 12:30:00 DESCRIPTION;LANGUAGE=en-US:We show how to transform any additively homomor phic private-key encryption scheme that is compact, into a public-key en cryption scheme. By compact we mean that the length of a homomorphically generated encryption is independent of the number of ciphertexts from which it was created. We do not require anything else on the distributi on of homomorphically generated encryptions (in particular, we do not re quire them to be distributed like real ciphertexts).

Our re sulting public-key scheme is homomorphic in the following sense. If the private-key scheme is $i+1$-hop homomorphic with respect to some set of operations then the public-key scheme we construct is $i$-hop homomorphi c with respect to the same set of operations. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400014550 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120104T133000 DTEND;TZID="Asia/Jerusalem":20120104T153000 DTSTAMP;TZID="Asia/Jerusalem":20120104T133000 FREEBUSY;FBTYPE=BUSY:20120104T133000/20120104T153000 SUMMARY;LANGUAGE=en-US:The msc talk by Yossi Attiya about Algorithmic cool ing of spins using optimal control at 2012-01-04 13:30:00 DESCRIPTION;LANGUAGE=en-US:Nuclear magnetic resonance (NMR) has proven to be a leading implementation of quantum information processors where each m olecule in the sample constitutes a register of quantum bits (qubits). How ever, at room temperature, the qubits that are realized by nuclear spins ( 1/2) are in a highly mixed state: noisy or with high entropy. Source codin g (compression) can cool some spins (reducing their Shannon entropy) while heating others, yet this closed-system technique is limited by Shannon's entropy conservation. Algorithmic cooling bypasses this bound by utilizing fast-reseting spins and thermalization, an interaction of the qubits with the environment, which is non-reversible. By using GRAPE, an optimal cont rol algorithm, we design efficient and robust radio frequency pulses, and applied multiple rounds of algorithmic cooling to a 3-spin system at the T echnion NMR lab. No previous knowledge of nuclear magnetic resonance, quan tum computers or data compression is required for the talk. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400014510 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120105T123000 DTEND;TZID="Asia/Jerusalem":20120105T143000 DTSTAMP;TZID="Asia/Jerusalem":20120105T123000 FREEBUSY;FBTYPE=BUSY:20120105T123000/20120105T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Zeev Dvir (Princeton Uni versity) about Theory Seminar: Subspace Evasive Sets at 2012-01-05 12:30:0 0 DESCRIPTION;LANGUAGE=en-US:We describe an explicit, simple, construction o f large subsets of F^n, where F is a finite field, that have small inter section with every k-dimensional affine subspace. Interest in the explic it construction of such sets, termed 'subspace-evasive' sets, started in the work of Pudlak and Rodl (2004) who showed how such constructions ov er the binary field can be used to construct explicit Ramsey graphs. Mor e recently, Guruswami (2011) showed that, over large finite fields (of size polynomial in n), subspace evasive sets can be used to obtain expl icit list-decodable codes with optimal rate and constant list-size.
We construct subspace evasive sets over large fields (polynomi al in n) and use them to derive the applications to list-decodable codes discovered by Guruswami. Our construction employs both combinatoric a nd algebraic methods.

(Joint work with Shachar Lovett, IAS) ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400014580 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120108T143000 DTEND;TZID="Asia/Jerusalem":20120108T163000 DTSTAMP;TZID="Asia/Jerusalem":20120108T143000 FREEBUSY;FBTYPE=BUSY:20120108T143000/20120108T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Keren Censor-Hillel about Fast D istributed Computing Despite Poor Connectivity at 2012-01-08 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014240 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120109T143000 DTEND;TZID="Asia/Jerusalem":20120109T163000 DTSTAMP;TZID="Asia/Jerusalem":20120109T143000 FREEBUSY;FBTYPE=BUSY:20120109T143000/20120109T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Roi Reichart about Efficient and Exact Inter-Sentence Decoding for Natural Language Processing at 2012-01- 09 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014130 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120109T183000 DTEND;TZID="Asia/Jerusalem":20120109T203000 DTSTAMP;TZID="Asia/Jerusalem":20120109T183000 FREEBUSY;FBTYPE=BUSY:20120109T183000/20120109T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Amir E. Aharoni abo ut Haifux Club: Maqaf Hataf Patakh - The new standard Hebrew Keyboard Layo ut at 2012-01-09 18:30:00 DESCRIPTION;LANGUAGE=en-US:In the last two years a committee in the Standa rds Institution of Israel worked, on my initiative and with my active part icipation, on a revision for the standard of the Hebrew keyboard layout. W hy did i want to change the layout? Why is it good to do it through the St andards Institution? How to make people who represent Microsoft, Linux, an d Apple agree about things? How is a keyboard layout edited in Linux and i n Windows? How do you install a keyboard layout and how do you distribute it to a lot of people? ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400014600 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120110T113000 DTEND;TZID="Asia/Jerusalem":20120110T133000 DTSTAMP;TZID="Asia/Jerusalem":20120110T113000 FREEBUSY;FBTYPE=BUSY:20120110T113000/20120110T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Margarita Osadchy (CS, Haifa University) about Pixel Club: Pose Estimation and Recognition using Spe cular Highlights at 2012-01-10 11:30:00 DESCRIPTION;LANGUAGE=en-US:We presents a novel approach to pose estimation and model-based recognition of specular objects in difficult viewing co nditions, such as low illumination, cluttered background, and large high lights and shadows that appear on the object of interest. In such challe nging conditions conventional features are unreliable. We show that unde r the assumption of a dominant light source, specular highlights produce d by a known object can be used to establish correspondence between its image and the 3D model, and to verify the hypothesized pose and the iden tity of the object.

Previous methods that use highlights f or recognition make limiting assumptions such as known pose, scene-depen dent calibration, simple shape, etc. The proposed method can efficiently recognize free form specular objects in arbitrary pose and under unknow n lighting direction. It uses only a single image of the object as its i nput and outputs object identity, full pose, and the lighting direction in the scene.

This is a joint work with Aaron Netz ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400014590 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120110T143000 DTEND;TZID="Asia/Jerusalem":20120110T163000 DTSTAMP;TZID="Asia/Jerusalem":20120110T143000 FREEBUSY;FBTYPE=BUSY:20120110T143000/20120110T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Or Zuk about Missing Heritabilit y: New algorithmic and statistical approaches at 2012-01-10 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014520 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120111T113000 DTEND;TZID="Asia/Jerusalem":20120111T133000 DTSTAMP;TZID="Asia/Jerusalem":20120111T113000 FREEBUSY;FBTYPE=BUSY:20120111T113000/20120111T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Gabi Nakibly (National EW Resear ch and Simulation Center) about ceClub: Persistent OSPF Attacks at 2012-01 -11 11:30:00 DESCRIPTION;LANGUAGE=en-US:Open Shortest Path First (OSPF) is the most pop ular interior gateway routing protocol on the Internet. Most known OSPF at tacks that have been published in the past are based on falsifying the lin k state advertisement (LSA) of an attacker-controlled router. These attack s can only falsify a small portion of the routing domain's topology, hence their effect is usually limited. More powerful attacks are the ones that affect LSAs of other routers not controlled by the attacker. However, thes e attacks usually trigger the OSPF "fight-back" mechanism by the victim ro uter which advertises a correcting LSA, making the attacks' effect non-per sistent. In this work we present new OSPF attacks that exploit design vuln erabilities in the protocol specification. These new attacks can affect th e LSAs of routers not controlled by the attacker while evading "fight-back ". As a result, an attacker can persistently falsify large portions of the routing domain's topology viewed by other routers thereby giving the atta cker control over their routing tables. We discuss a number of mitigation strategies and propose an update to the OSPF specification that defeats th ese attacks and improves OSPF security. The talk is based on results prese nted at Black Hat '11 and NDSS '12. This is a joint work with Alex Kirshon , Dima Gonikman and Dan Boneh. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400014620 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120111T123000 DTEND;TZID="Asia/Jerusalem":20120111T143000 DTSTAMP;TZID="Asia/Jerusalem":20120111T123000 FREEBUSY;FBTYPE=BUSY:20120111T123000/20120111T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Brent Waters (University of Texas) about Theory Seminar: Detecting Dangerous Queries: A New Approa ch for Chosen Ciphertext Security at 2012-01-11 12:30:00 DESCRIPTION;LANGUAGE=en-US:I will present a new approach for creating chos en ciphertext secure encryption. The focal point of our work is a new a bstraction that we call Detectable Chosen Ciphertext Security (DCCA). I ntuitively, this notion is meant to capture systems that are not necess arily chosen ciphertext attack (CCA) secure, but where we can detect wh ether a certain query CT can be useful for decrypting (or distinguishing) a challenge ciphertext CT*.

We show how to build chose n ciphertext secure systems from DCCA security. We motivate our techniq ues by describing multiple examples of DCCA systems including creating the m from 1-bit CCA secure encryption - capturing the recent Myers-shelat result (FOCS 2009). Our work identifies DCCA as a new target for buildi ng CCA secure systems.

(This is joint work with Susan Hoh enberger and Allison Lewko.) ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400014610 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120111T143000 DTEND;TZID="Asia/Jerusalem":20120111T163000 DTSTAMP;TZID="Asia/Jerusalem":20120111T143000 FREEBUSY;FBTYPE=BUSY:20120111T143000/20120111T163000 SUMMARY;LANGUAGE=en-US:The phd talk by Tommer Kotek about Definability of Combinatorial Functions at 2012-01-11 14:30:00 DESCRIPTION;LANGUAGE=en-US:The complexity of several prominent graph polyn omials, such as the chromatic polynomial and the matching polynomial, has been studied in the literature. In 2008 Makowsky raised a conjecture wh ich generalizes complexity results for specific graph polynomials to an i nfinite class of graph polynomials which include almost all of those in th e literature. The conjecture states roughly that the evaluations of such graph polynomials are equivalent in terms of running-time complexity, e xcept for a small and nicely structured exception set. As a case stu dy for the conjecture, we consider a graph polynomial which originated in statistical mechanics in order to study phase transitions. A trivariate Ising polynomial was studied with respect to its approximability by L. A . Goldberg, M. Jerrum and M. Paterson in 2003. A related bivariate Ising polynomial was studied in by D. Andr\'{e}n and K. Markstr\"{o}m in 2009 . We show that both Ising polynomials satisfy versions of Makowsky's con jecture, even when the domain is restricted. Under a counting version of the Exponential Time Hypothesis we give a dichotomy theorem stating that evaluations of the bivariate Ising polynomial either require exponential time runnning-time, or can be computed in polynomial time. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012210 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120111T143000 DTEND;TZID="Asia/Jerusalem":20120111T163000 DTSTAMP;TZID="Asia/Jerusalem":20120111T143000 FREEBUSY;FBTYPE=BUSY:20120111T143000/20120111T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Joseph Keshet about Making Compu ters Good Listeners at 2012-01-11 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014570 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120117T113000 DTEND;TZID="Asia/Jerusalem":20120117T133000 DTSTAMP;TZID="Asia/Jerusalem":20120117T113000 FREEBUSY;FBTYPE=BUSY:20120117T113000/20120117T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Michal Jacob (Intel, Compute r Vision Group) about Pixel Club: The Role of Target Fixations in the Proc ess of Recognition at 2012-01-17 11:30:00 DESCRIPTION;LANGUAGE=en-US:Why do we perceive some elements in a visual sc ene, while others remain undetected? We compared fixations on detected v s. undetected items in the Identity Search Task (Jacob & Hochstein, 2009 ). Using a gaze-contingent technique, we further controlled the number o f fixations on the target (Jacob & Hochstein, 2010). Results show that d etected targets were fixated at a greater extent, and a backward dynamic s alignment revealed a bifurcation point where the differential characte ristics begin. Moreover, the greater the number of target fixations, the better the recognition, as manifested in a decrease in response time, i ncreases in hit-rate and detectability, d', and increase in reported res ponse confidence. The results suggest that target fixations lead to an e arly implicit recognition which in turn leads to more fixations, and ult imately to full explicit recognition. We constructed a model (Jacob & Ho chstein, 2011) which simulates the relative information that is availabl e to the observer for each scene unit, at any given moment. Depending on the eye movement scan-path, the available information is incremented by the fixations as well as affected by memory decay. The model results re flect the experimental conclusion that several target fixations are need ed for processing visual information in order to achieve recognition. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014640 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120117T143000 DTEND;TZID="Asia/Jerusalem":20120117T163000 DTSTAMP;TZID="Asia/Jerusalem":20120117T143000 FREEBUSY;FBTYPE=BUSY:20120117T143000/20120117T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Amit Singer about Vector Diffusi on Maps and the Connection Laplacian at 2012-01-17 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014630 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120118T113000 DTEND;TZID="Asia/Jerusalem":20120118T133000 DTSTAMP;TZID="Asia/Jerusalem":20120118T113000 FREEBUSY;FBTYPE=BUSY:20120118T113000/20120118T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Stacy Patterson (EE, Technion) a bout ceClub: Distributed Average Consensus and Coherence in Dynamic Networ ks at 2012-01-18 11:30:00 DESCRIPTION;LANGUAGE=en-US:In the distributed average consensus problem, e ach node in a network has an initial value, and the objective is for all n odes to reach consensus at the average of these values using only communic ation with nearby nodes. Distributed average consensus algorithms have a w ide variety of applications, including distributed optimization, sensor fu sion, load balancing, and autonomous vehicle formation control.

This talk centers on the analysis of distributed averaging algorithms for consensus and vehicle formation control in networks with dynamic chara cteristics such as stochastic packet loss, node failures, network partitio ns, and additive disturbances. I will present an overview of distribued av eraging algorithms and some recent results on the stability and robustness of these algorithms in dynamic networks. We analyze the relationship betw een algorithm performance and the size, dimension, and dynamic characteris tics of the network, and we show that network dimension imposes fundamenta l limitations on algorithm performance in large networks.

bi o: Dr. Stacy Patterson received her B.A in Mathematics and B.S. in Compute r Science from Rutgers University in 1998 and her M.S. and Ph.D. in Comput er Science from the University of California, Santa Barbara in 2003 and 20 09 respectively. From July 2009 to August 2011, she was a postdoctoral sch olar at the Center for Control, Dynamical Systems, and Computation at the University of California, Santa Barbara. She is currently a postdoctoral f ellow in the Department of Electrical Engineering at the Technion, working with Professor Idit Keidar. Her research interests include distributed sy stems, vehicle networks, and cloud computing. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 861 UID:111ec4410400014660 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120118T123000 DTEND;TZID="Asia/Jerusalem":20120118T143000 DTSTAMP;TZID="Asia/Jerusalem":20120118T123000 FREEBUSY;FBTYPE=BUSY:20120118T123000/20120118T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Niv Buchbinder (Open Uni versity ) about Theory Seminar: A Polylogarithmic-Competitive Algorithm fo r the $k$-Server Problem at 2012-01-18 12:30:00 DESCRIPTION;LANGUAGE=en-US:The $k$-server problem is one of the most funda mental and extensively studied problems in online computation. Suppose the re is an $n$-point metric space and $k$ servers are located at some of the points of the metric space. At each time step, an online algorithm is giv en a request at one of the points of the metric space, and this request is served by moving a server to the requested point (if there is no server t here already). The cost of serving a request is defined to be the distance traveled by the server. Given a sequence of requests, the task is to devi se an online strategy minimizing the sum of the costs of serving the reque sts.

We give the first polylogarithmic-competitive randomized online algorithm for the $k$-server problem on an arbitrary finite metric space. In particular, our algorithm achieves a competitive ratio of $O(\l og^3 n \log^2 k)$ for any metric space on $n$ points. Our algorithm improv es upon the deterministic $(2k-1)$-competitive algorithm of Koutsoupias an d Papadimitriou whenever $n$ is sub-exponential in $k$.

[This paper won the last FOCS best paper award] ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400014650 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120122T123000 DTEND;TZID="Asia/Jerusalem":20120122T143000 DTSTAMP;TZID="Asia/Jerusalem":20120122T123000 FREEBUSY;FBTYPE=BUSY:20120122T123000/20120122T143000 SUMMARY;LANGUAGE=en-US:The CSpecial Talk talk by David Carmel about at 20 12-01-22 12:30:00 DESCRIPTION;LANGUAGE=en-US:David Carmel, from IBM Research, Haifa, wil tal k about the IBM Watson project. The talk is given in the course 'Introduc tion to Artificial Intelligence", but is open to the public. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 3 UID:111ec4410400014720 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120124T113000 DTEND;TZID="Asia/Jerusalem":20120124T133000 DTSTAMP;TZID="Asia/Jerusalem":20120124T113000 FREEBUSY;FBTYPE=BUSY:20120124T113000/20120124T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Dana Segev (EE, Technion) ab out Pixel Club: Visual Audio Denoising at 2012-01-24 11:30:00 DESCRIPTION;LANGUAGE=en-US:Audio denoising is a long studied problem, with numerous algorithms and a wide accumulated knowledge. Considering non-s tationary noise (like noise in a cocktail party environment) and strong, this task becomes very difficult to handle. This paper considers such a udio denoising problems, where the audio track is accompanied by a video . Furthermore, the disturbing source is generally not visible in the fie ld of view, and its nature is unknown. Overcoming such unknown noise is different than source separation, where the disturbing sources or their priors are at least partially accessible. We demonstrate the core-abilit y to use the video information to better filter the audio. The approach we take in this work is an example-based one, assuming that we have at o ur disposal a relatively good-quality movie (video+audio) to train on. T he noise removal itself is done by processing short temporal segments of video-and-audio, seeking relevant examples from the training set. The v ideo information eliminates non- relevant training examples and improves the algorithm performance. We demonstrate this approach on two differen t types of signals. In all the experiments, a very significant denoising is achieved, also in cases where audio-only processing methods fail.

MSc thesis under the supervision of Michael Elad and Yoav Schec hner. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400014700 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120125T123000 DTEND;TZID="Asia/Jerusalem":20120125T143000 DTSTAMP;TZID="Asia/Jerusalem":20120125T123000 FREEBUSY;FBTYPE=BUSY:20120125T123000/20120125T143000 SUMMARY;LANGUAGE=en-US:The msc talk by Daniel Hurwitz about Exploiting Mor phology in Identifying Multiword Expressions from Multilingual Parallel Co rpora at 2012-01-25 12:30:00 DESCRIPTION;LANGUAGE=en-US:Our research discusses multi-word expressions ( MWE) such as "kick the bucket", "hot dog", "by and large", "look up, and " spill the beans". Identification of MWEs has been a hot subject of researc h in recent years. We present a method to identify MWEs in Hebrew using a new concept we term Language Isolation. We focus on dissecting (or "isolat ing") the morphological properties of words to discover potential MWEs and show that this method improves the alignment of multi-lingual (Spanish-En glish-Hebrew) parallel corprora. We then use our method to acquire Hebrew MWEs. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400014690 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120125T123000 DTEND;TZID="Asia/Jerusalem":20120125T143000 DTSTAMP;TZID="Asia/Jerusalem":20120125T123000 FREEBUSY;FBTYPE=BUSY:20120125T123000/20120125T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Eran Omri (Bar-Ilan Univ ersity) about Theory Seminar: Coin Flipping with Constant Bias Implies One -Way Functions at 2012-01-25 12:30:00 DESCRIPTION;LANGUAGE=en-US:It is well known (c.f., Impagliazzo and Luby [F OCS '89]) that the existence of almost all ``interesting" cryptographic applications, i,e., ones that cannot hold information theoretically, imp lies one-way functions. An important exception where the above implicati on is not known, however, is the case of coin-flipping protocols. Such p rotocols allow honest parties to mutually flip an unbiased coin, while guaranteeing that even a cheating (efficient) party cannot bias the out put of the protocol by much. While Impagliazzo and Luby proved that coin -flipping protocols that are safe against negligible bias do imply one-w ay functions, and, very recently, Maji, Prabhakaran, and Sahai [FOCS '10 ] proved the same for constant-round protocols (with any non-trivial bia s). For the general case, however, no such implication was known.

We make progress towards answering the above fundamental question, showing that (strong) coin-flipping protocols safe against a constant bias (concretely, \frac{\sqrt2 -1}2 - o(1)) imply one-way functions. Joi nt work with Iftach Haitner. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400014710 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120131T113000 DTEND;TZID="Asia/Jerusalem":20120131T133000 DTSTAMP;TZID="Asia/Jerusalem":20120131T113000 FREEBUSY;FBTYPE=BUSY:20120131T113000/20120131T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Greg Shakhnarovich (TTI-Chic ago) about Pixel Club: Diverse M-best Solutions in MRFs at 2012-01-31 11:3 0:00 DESCRIPTION;LANGUAGE=en-US:Much effort has been directed at algorithms for obtaining the highest probability (MAP) configuration in a probabilisti c (random field) model. In many situations, one could benefit from addit ional solutions with high probability. Current methods for computing add itional most probable configurations produce solutions that tend to be v ery similar to the MAP solution and each other. This is often an undesir able property. I will describe an algorithm for the M-Best Mode problem, which involves finding a diverse set of highly probable solutions under a discrete probabilistic model. Given a dissimilarity function measurin g difference between two solutions, our algorithm maximizes a linear com bination of the probability and dissimilarity to previous solutions. Thi s is a generalization of the M-Best MAP problem and we show that for cer tain families of dissimilarity functions we can guarantee that these sol utions can be found as easily as the MAP solution.

Joint work with Payman Yadollahpour and Dhruv Batra. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014750 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120131T143000 DTEND;TZID="Asia/Jerusalem":20120131T163000 DTSTAMP;TZID="Asia/Jerusalem":20120131T143000 FREEBUSY;FBTYPE=BUSY:20120131T143000/20120131T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Ouri Wolfson about How Your Smar tphone Can Use Game-Theory to Valet-park Your Car at 2012-01-31 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014670 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120131T153000 DTEND;TZID="Asia/Jerusalem":20120131T173000 DTSTAMP;TZID="Asia/Jerusalem":20120131T153000 FREEBUSY;FBTYPE=BUSY:20120131T153000/20120131T173000 SUMMARY;LANGUAGE=en-US:The colloq talk by Ran Ginosar about The Plural Arc hitecture: Shared Memory Many-cores with Hardware Scheduling at 2012-01-31 15:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014680 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120201T110000 DTEND;TZID="Asia/Jerusalem":20120201T130000 DTSTAMP;TZID="Asia/Jerusalem":20120201T110000 FREEBUSY;FBTYPE=BUSY:20120201T110000/20120201T130000 SUMMARY;LANGUAGE=en-US:The phd talk by Alex Kogan about Utilizing Multiple Radio Interfaces in Wireless Networks at 2012-02-01 11:00:00 DESCRIPTION;LANGUAGE=en-US:Contemporary mobile devices are equipped with m ultiple wireless interfaces, such as WiFi, Bluetooth, WiMax, ZigBee, NF C, etc. All these technologies differ dramatically one from another in max imum transmission range, bandwidth and power demands. Among all subsyst ems operating inside mobile devices, wireless communication is known as being particularly power-hungry, accounting for 50-70% of the total po wer consumption in small handheld devices, such as smartphones, and for 10 % in laptops. Given the varying characteristics of different wireless t echnologies, an obvious question arises: Can we utilize the presence of multiple interfaces on contemporary mobile devices in order to improve their wireless networking capabilities in general and power efficiency in particular? In this thesis, we investigate this question from multip le perspectives: theoretical, more practical and fully experimental. On the theoretical side, we formulate a novel optimization problem. A solu tion to this problem defines a network topology where some devices turn of f their power-hungry interface while every device still remains connect ed to the rest of the network by at least one interface. Through theoret ical analysis and practical implementation, we show that the proposed appr oach achieves significant energy savings, which increase with the size ( and the density) of the network. On the more experimental side of our wo rk, we perform a combined power and throughput performance study of WiFi and Bluetooth in smartphones. In the process, we discover several interes ting phenomena, some of which counter previous conventions, and draw so me operative suggestions for researchers and smartphone developers. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400012190 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120206T143000 DTEND;TZID="Asia/Jerusalem":20120206T163000 DTSTAMP;TZID="Asia/Jerusalem":20120206T143000 FREEBUSY;FBTYPE=BUSY:20120206T143000/20120206T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Haim Avron SPECIAL GUEST TALK- a bout Random Sampling Preconditioners at 2012-02-06 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014770 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120206T183000 DTEND;TZID="Asia/Jerusalem":20120206T203000 DTSTAMP;TZID="Asia/Jerusalem":20120206T183000 FREEBUSY;FBTYPE=BUSY:20120206T183000/20120206T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Avi Mendelson (Micr osoft and CS&EE, Technion) about Haifux Club: C++ AMP: Microsoft approach for Heterogeneous Computing at 2012-02-06 18:30:00 DESCRIPTION;LANGUAGE=en-US:C++ AMP (Accelerated Massive Parallelism) is a native programming model that contains elements that span the C++ programm ing language and its runtime library. The syntactic changes introduced by AMP are minimal, but additional restrictions are enforced to reflect the l imitations of data parallel hardware. Data parallel algorithms are support ed by the introduction of multi-dimensional array types, array operations on those types, indexing, asynchronous memory transfer, shared memory, syn chronization and tiling/partitioning techniques.

The language aims to support heterogeneous programming environments, starting from GPG PU and up-to Cloud computing. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400014800 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120207T113000 DTEND;TZID="Asia/Jerusalem":20120207T133000 DTSTAMP;TZID="Asia/Jerusalem":20120207T113000 FREEBUSY;FBTYPE=BUSY:20120207T113000/20120207T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Moti Freiman (Computational Radiology Lab, Harvard Medical School) about Pixel Club: Tissue Microenvi ronment Magnetic Resonance Imaging (TM-MRI) of the body: A Reliable Quant itative Biomarker for Personalized Treatment Paradigms at 2012-02-07 11:30 :00 DESCRIPTION;LANGUAGE=en-US:Personalized treatment approaches which optimiz e drugs doses according to pre-treatment and early response-to-therapy eva luation hold the promise to improve treatment success rates and reduce sev ere adverse side-effects due to drugs toxicity in variety of pathologies. Reliable assessment of tissue microenvironment including cell proliferatio n, density and size and tissue perfusion as a biomarker for disease activi ty is a key necessity for personalized, response-based treatment regimes. Histology-based tissue microenvironment analysis requires invasive, surg ical procedure to obtain the tissue sample. Moreover, the histological ana lysis is limited to the obtained tissue sample which may not be sufficient in heterogeneous microenvironment. Instead, we developed the TM-MRI metho d utilized short-duration free-breathing diffusion-weighted MRI acquired w ith multiple b-values coupled with global tissue microenvironment model an d reliable fitting technique that enables radiation and toxicity-free non- invasive insight into the entire three-dimensional tissue microstructure. In the lecture I’ll present the core components of this method: 1) the TM-MRI image acquisition scheme; 2) the global tissue microenvironment mo del, and; 3) reliable model fitting technique with intrinsic fit-quality a ssessment. In addition, I'll present initial clinical results of non-invas ive disease activity assessment using the TM-MRI quantitative biomarkers i n pediatric Crohn’s disease patients and discuss future applications of this technique. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400014790 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120207T143000 DTEND;TZID="Asia/Jerusalem":20120207T163000 DTSTAMP;TZID="Asia/Jerusalem":20120207T143000 FREEBUSY;FBTYPE=BUSY:20120207T143000/20120207T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Ivan Damgaard about Multiparty C omputation from Somewhat Homomorphic Encryption at 2012-02-07 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014780 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120208T113000 DTEND;TZID="Asia/Jerusalem":20120208T133000 DTSTAMP;TZID="Asia/Jerusalem":20120208T113000 FREEBUSY;FBTYPE=BUSY:20120208T113000/20120208T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Roman Manevich (UT Austin) about ceClub: Synthesizing Concurrent Relational Data Structures at 2012-02-08 11:30:00 DESCRIPTION;LANGUAGE=en-US:Efficient concurrent data structures are extrem ely important for obtaining good performance for most parallel programs. H owever, ensuring the correctness of concurrent data structure implementati ons can be very tricky because of concurrency bugs such as race conditions and deadlocks. In systems that use optimistic parallel execution such as boosted transactional memory systems and the Galois system, the implementa tion of concurrent data structures is even more complex because data struc ture implementations must also detect conflicts between concurrent activit ies and support the rollback of conflicting activities.

At pr esent, these types of concurrent data structures are implemented manually by expert programmers who write explicitly parallel code packaged into lib raries for use by application programmers. This solution has its limitatio ns; for example, it does not permit the customization or tuning of a data structure implementation for a particular application.

In thi s talk, we present Autograph, which is the first concurrent data structure compiler that can synthesize concurrent relational data structure impleme ntations for use by application programmers. The input to Autograph is a h igh-level declarative specification of an abstract data type (ADT); the ou tput is a concurrent implementation of that ADT with conflict detection an d rollback baked in. Our synthesizer is parameterized by a set of data str uctures called "tiles", which are building blocks that the compiler compos es to create the overall data structure. Application programmers can use a simple expression language to tune the composition of these tiles, thereb y exercising high level, fine-grain control of data structure implementati ons. We have used Autograph to synthesize concurrent sparse graph data str uctures for a number of complex parallel graph benchmarks. Our results sho w that the synthesized concurrent data structures usually perform better t han the handwritten ones; for some applications and thread counts, they im prove performance by a factor of 2. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 861 UID:111ec4410400014760 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120208T130000 DTEND;TZID="Asia/Jerusalem":20120208T150000 DTSTAMP;TZID="Asia/Jerusalem":20120208T130000 FREEBUSY;FBTYPE=BUSY:20120208T130000/20120208T150000 SUMMARY;LANGUAGE=en-US:The phd talk by Guy Sagy about Geometric Methods fo r Analyzing and Monitoring Large Distributed Data at 2012-02-08 13:00:00 DESCRIPTION;LANGUAGE=en-US:A basic requirement in many distributed systems is the ability to detect objects whose score, according to a given funct ion, exceeds some threshold. Since an object's data can be partitioned ov er various nodes, computing its global score requires collecting its data over the network. A main challenge is to perform threshold queries or m onitoring with minimum network communication, i.e., without collecting th e data from the nodes to a central location. In this talk I will presen t an application of geometric ideas for performing threshold queries and top-k queries over distributed databases with minimum communication. The proposed method can handle general functions by representing them as a d ifference of monotonic functions, and imposing thresholds on the latter. In addition, I will present a novel scheme for communication reduction in distributed monitoring using local constraints. Communication is req uired only in the event that the local constraints are violated by the in coming data. As opposed to previous work on geometric monitoring, here t he local constraints are tailored to fit the local data distribution at e ach stream, and satisfy an optimality criterion. The result is a substant ial decrease in the required volume of communication compared to previous state of the art, up to two orders of magnitude in experiments with real -life data. Theoretical analysis suggests that the reduction can sometime s be unbounded, and that it typically improves with the dimensionality of the data, which is borne out in the experiments. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400014740 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120214T143000 DTEND;TZID="Asia/Jerusalem":20120214T163000 DTSTAMP;TZID="Asia/Jerusalem":20120214T143000 FREEBUSY;FBTYPE=BUSY:20120214T143000/20120214T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Roman Manevich about Synthesizin g Concurrent Relational Data Structures at 2012-02-14 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014730 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120219T113000 DTEND;TZID="Asia/Jerusalem":20120219T133000 DTSTAMP;TZID="Asia/Jerusalem":20120219T113000 FREEBUSY;FBTYPE=BUSY:20120219T113000/20120219T133000 SUMMARY;LANGUAGE=en-US:The msc talk by Itamar Hartstein about On the Compl exity of the Regenerator Location Problem - Treewidth and Other Parameters at 2012-02-19 11:30:00 DESCRIPTION;LANGUAGE=en-US:We deal with the Regenerator Location Problem i n optical networks. We are given a network G = (V, E), and a set Q of comm unication requests between pairs of terminals in V. We investigate two var iations: one in which we are given a routing P of the requests in Q, and o ne in which we are required to find also the routing. In both cases, each path in P must contain a regenerator after every d edges in order to deal with loss of signal quality for some d > 0. The goal is to minimize the nu mber of vertices that contain regenerators used by the solution. Both vari ations of the problem are NP-Hard in the general case. In this work we inv estigate the parameterized complexity of the problem. We introduce several fixed parameter tractability results and polynomial algorithms for fixed parameter values, as well as several NP-Hardness results. The main paramet ers under consideration are the treewidth of the input graph, and the numb er of connections. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400014830 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120220T183000 DTEND;TZID="Asia/Jerusalem":20120220T203000 DTSTAMP;TZID="Asia/Jerusalem":20120220T183000 FREEBUSY;FBTYPE=BUSY:20120220T183000/20120220T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Gai Shaked about Ha ifux Club: Making Hebrew Slides with LaTeX and Beamer at 2012-02-20 18:30 :00 DESCRIPTION;LANGUAGE=en-US:LaTeX is a high-quality typesetting system; BEA MER is is a LaTeX class for creating presentations; Hebrew is an esoteric west Semitic language we happen to speak. This talk will cover all three, by the end of the talk you will be able to create beautiful and comprehens ive Hebrew presentations using Beamer in minutes.

The talk a ssumes prior knowledge in Hebrew, but not in LaTeX nor Beamer. We will cov er basic concepts, commands and examples in depth, and finally add some po inters to advanced resources. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400014890 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120221T113000 DTEND;TZID="Asia/Jerusalem":20120221T133000 DTSTAMP;TZID="Asia/Jerusalem":20120221T113000 FREEBUSY;FBTYPE=BUSY:20120221T113000/20120221T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Leah Bar (Mathematics, Tel A viv University) about Pixel Club: Hierarchical Invariant Sparse Modeling f or Image Analysis at 2012-02-21 11:30:00 DESCRIPTION;LANGUAGE=en-US:Sparse representation theory has been increasin gly used in signal processing and machine learning. In this work we intr oduce a hierarchical sparse modeling approach which integrates information from the image patch level to derive a mid-level invariant image and patt ern representation. The proposed framework is based on a hierarchical arch itecture of dictionary learning for sparse coding in a cortical (log-polar ) space, combined with a novel pooling operator which incorporates the Rap id transform and max pooling to attain rotation and scale invariance. The invariant sparse representation of patterns here presented- can be use d in different object recognition tasks. Promising results are obtained for three applications -- 2D shapes classification, texture recognition and object detection.

joint work with Guillermo Sapiro, Unive rsity of Minnesota. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400014820 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120228T143000 DTEND;TZID="Asia/Jerusalem":20120228T163000 DTSTAMP;TZID="Asia/Jerusalem":20120228T143000 FREEBUSY;FBTYPE=BUSY:20120228T143000/20120228T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Ido Tal about Polar codes: const ruction and improved decoding at 2012-02-28 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014810 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120229T143000 DTEND;TZID="Asia/Jerusalem":20120229T163000 DTSTAMP;TZID="Asia/Jerusalem":20120229T143000 FREEBUSY;FBTYPE=BUSY:20120229T143000/20120229T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Nati Srebro about Matrix Learnin g: A Tale of Two Norms at 2012-02-29 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014840 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120306T140000 DTEND;TZID="Asia/Jerusalem":20120306T160000 DTSTAMP;TZID="Asia/Jerusalem":20120306T140000 FREEBUSY;FBTYPE=BUSY:20120306T140000/20120306T160000 SUMMARY;LANGUAGE=en-US:The msc talk by Assaf Rappaport about Approximation Algorithm for Soft-Capacitated Connected Facility at 2012-03-06 14:00:00 DESCRIPTION;LANGUAGE=en-US:Data centers are becoming the hosting platform for a wide spectrum of composite applications. In recent years, large inv estments have been made in massive data centers supporting cloud services , by companies such as eBay, Facebook, Google, Microsoft, and Yahoo!. Wit h an increasing trend towards global and more communication intensive ap plications, the bandwidth usage within and between data centers is rapidl y growing. The placement of the data used by these global applications pr esents a challenging optimization problem, involving several factors. We study a network design problem that combines facility location and conne ctivity problems. Consider an email application that uses on an authentic ation service, and consider the problem of placing replicas of an object (e.g., the authentication service) at multiple locations in the Cloud. Re plica placement deals with the actual number and network location of the replicas. Clearly, we would like to minimize the network distance between an application server in a data center and the closest replica containin g the desired content and thus having more replicas helps. On the other h and, having more replicas is more expensive so we need to model the cost and the benefit in a way that will allow us to make the appropriate deci sions regarding the number and the network locations of the replicas. Th is problem is strongly related to a family of optimization problems gener ally referred to as facility location problems. As most of the existing a lgorithms neglect the cost of keeping the replicas across the network up to date, we believe that considering this factor can lead to better reali stic solutions. A replica must be synchronized with the original content server in order to provide reliable and precise response to the client re quests. The synchronization traffic across the network depends on the num ber of replicas deployed in the network, the topology of the distributed update and the rate of updates in the content of the server. Moreover, we extend our model by adding capacity constraint for each replica. We can model the scenario above as a Soft-Capacitated Connected Facility Locatio n Problem which is NP-Hard in the general case. In this work we introduc e a constant approximation algorithm for this problem and study its appli cability in the Cloud data placement paradigm. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400014950 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120307T143000 DTEND;TZID="Asia/Jerusalem":20120307T163000 DTSTAMP;TZID="Asia/Jerusalem":20120307T143000 FREEBUSY;FBTYPE=BUSY:20120307T143000/20120307T163000 SUMMARY;LANGUAGE=en-US:The msc talk by Tomer Koren about Learning Linear S upport Vector Machines in Sublinear Time at 2012-03-07 14:30:00 DESCRIPTION;LANGUAGE=en-US:In recent years, stochastic approximation appro aches such as Stochastic Gradient Descent (SGD) and Stochastic Dual Avera ging have become the optimization method of choice for many learning prob lems, including linear Support Vector Machines (SVMs). This is not surpri sing, since such methods yield optimal generalization guarantees with onl y a single pass over the data. They in a sense have an unbeatable runtime : the runtime to get to a desired generalization goal is the same as the size of the data set required to do so. In this work we show, for the first time, how to beat this unbeatable runtime, and present a method th at in a certain relevant regime, learns in runtime less than the size of the minimal training set size required for learning. Our method is based on a stochastic primal-dual approach, where the primal step is akin to an importance-weighted SGD, and the dual step is a stochastic update on the importance weights. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400014930 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120312T140000 DTEND;TZID="Asia/Jerusalem":20120312T160000 DTSTAMP;TZID="Asia/Jerusalem":20120312T140000 FREEBUSY;FBTYPE=BUSY:20120312T140000/20120312T160000 SUMMARY;LANGUAGE=en-US:The msc talk by Elad Gidron about NUMA and locality aware multi-core software at 2012-03-12 14:00:00 DESCRIPTION;LANGUAGE=en-US:Emerging computer architectures pose many new c hallenges for software development. First, as the number of computing ele ments constantly increases, the importance of scalability of parallel prog rams becomes more significant. Second, accessing memory has become the p rincipal bottleneck, while multi-CPU systems are based on NUMA architectu res, where memory access from different chips is asymmetric. Therefore, it is important to design software with local data access, cache-friendlin ess, and reduced contention on shared memory locations, especially acros s chips. In our work we focus on two problems: 1. We design and imp lement a scalable and highly-efficient non-blocking consumer-producer task pool, with lightweight synchronization-free operations in the common case. Its data allocation scheme is cache-friendly and highly suitab le for NUMA environments. Moreover, our pool is robust in the face of imbalanced loads and unexpected thread stalls. 2. We consider the case o f improving metadata locality in word-based STMs. To this end, we evaluate a locality-conscious approach for maintaining versioned locks in TL2 . The speedup of the improved algorithm reaches a hundred percent on STAMP benchmarks. We show that this speedup stems from the following factors: 1) improved spacial and temporal locality, 2) reduced false shari ng and 3) less false conflicts. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400014990 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120313T143000 DTEND;TZID="Asia/Jerusalem":20120313T163000 DTSTAMP;TZID="Asia/Jerusalem":20120313T143000 FREEBUSY;FBTYPE=BUSY:20120313T143000/20120313T163000 SUMMARY;LANGUAGE=en-US:The TCE Lecture talk by Prof. Theo Ungerer (CS, Uni versity of Augsberg, Germany) about TCE Special Guest Lecture: Obstacles and Chances for Multi-core Deployment in Hard Real-time Systems at 2012-03 -13 14:30:00 DESCRIPTION;LANGUAGE=en-US:Providing higher performance than state-of-the- art embedded processors can deliver today will increase safety, comfort, n umber and quality of services, while also lowering emissions as well as fu el demands for automotive, avionic and automation applications. Engineers who design hard real-time embedded systems in such embedded domains expres s a need for several times the performance available today while keeping s afety as major criterion. A breakthrough in performance is expected by par allelising hard real-time applications and running them on an embedded mul ti-core processor, which enables combining the requirements for high-perfo rmance with time-predictable execution.

The talk will discus s results of the EC FP-7 project MERASA (Multi-Core Execution of Hard Real -Time Applications Supporting Analysability, 2007-2011) and objectives of the parMERASA project (Multi-Core Execution of Parallelised Hard Real-Time Applications Supporting Analysability, 2011-2014). Both projects target t iming analysable systems of parallel hard real-time applications running o n a scalable multi-core processor. MERASA delivered a fully timing analysa ble four-core SMT processor as FPGA prototype together with adapted system software and WCET tools, running a parallelised version of a Honeywell In ternational autonomous flying vehicle code as demonstrator. parMERASA shif ts its objectives even more towards parallelisation of hard real-time appl ication software. To this end application companies of avionics, automotiv e, and construction machinery domains cooperate with tool developers and m ulti-core architects. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 861 UID:111ec4410400015050 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120314T113000 DTEND;TZID="Asia/Jerusalem":20120314T133000 DTSTAMP;TZID="Asia/Jerusalem":20120314T113000 FREEBUSY;FBTYPE=BUSY:20120314T113000/20120314T133000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Michael G. Katze ( University of Washington, Seattle) about Bioinformatics Forum: Systems Bio logy of Viral Pathogenesis and Immunity: Where are Google and IBM? at 2012 -03-14 11:30:00 DESCRIPTION;LANGUAGE=en-US:After decades of research, vaccines against som e of the greatest viral threats are still lacking and antiviral drugs r emain few and slow in coming. These shortcomings point to the need for n ew approaches that go beyond traditional virology methods. High-through put technologies and computational biology promise to deliver a much-nee ded boost to the field. My laboratory is using systems biology and compu tational approaches to understand and model integrated views of virus-ho st interactions, viral evasion of host defenses, and viral pathogenesis. Much of our work is focused on viruses responsible for worldwide pandem ics, including influenza virus, hepatitis C virus, and human immunodefic iency virus. As new experimental systems and technologies continue to co me online, such as mouse systems genetics, metabolomics, lipidomics, and next-generation sequencing, our systems-level views have expanded to en compass host genetic variation, metabolic pathways, epigenetics, microRN As, and long noncoding RNAs. Because this amount of information is beyon d the capacity of human intuition to grasp, mathematical frameworks and computational models must be constructed. Such models are necessary to p redict how molecular components work together to yield operational mecha nisms and phenotypic outcome. Models also provide predictions for how to most effectively design new diagnostics, therapeutics, and vaccines. Th e combination of high-throughput datasets and computational methods provid es best hope for understanding viral pathogenesis and speeding vaccine a nd drug development.

Michael Katze is a Professor of Microb iology at the University of Washington and Associate Director and Core S taff Scientist at the Washington National Primate Research Center. He is also Director of the Center for Systems and Translational Research on I nfectious Disease (STRIDE). He has studied virus-host interactions for m ore than 30 years and is a world leader in the use of systems biology ap proaches, including high-throughput technologies and computational metho ds, to understand, define and model virus-host interactions in a broad r ange of experimental systems. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 401 UID:111ec4410400014850 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120314T123000 DTEND;TZID="Asia/Jerusalem":20120314T143000 DTSTAMP;TZID="Asia/Jerusalem":20120314T123000 FREEBUSY;FBTYPE=BUSY:20120314T123000/20120314T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Ido Ben-Zvi (EE, Technio n) about Theory Seminar: Causality, Knowledge and Coordination in Synchron ous Systems at 2012-03-14 12:30:00 DESCRIPTION;LANGUAGE=en-US:Coordinating the proper ordering of events acro ss remote sites is a central task of distributed applications. In async hronous systems, such coordination depends in an essential way upon messag e chains, as captured by Lamport's happened-before relation. The relatio n provides a useful approximation of causality, in the sense that in asy nchronous systems two event can only be causally related if they are Lam port related.

The talk will consider coordination and causali ty in synchronous systems, where upper bounds are available on message tr ansmission times over each channel, and processes share a global clock. He re, communication is still essential whenever a spontaneous external input is meant to trigger an ordered sequence of responses across various sites . We capture the essence of such a coordination task in a proposed class o f coordination problems called Ordered Response. Within this framework we embark on a search for a similar notion of causality. We will not ouch upo n knowledge in any great depth, but consider that in a synchronous setting both message chains and the passage of time can be used to spread informa tion across the system, and hence to enable coordination.

Ind eed, it turns out that the synchronous analog for Lamport's causality is a structure that carefully combines message chains and the existing upper bounds on transmission times. This causal structure, called the centipede , is shown to be necessary in every solution of the Ordered Response probl em.

Time permitting, we will also discuss the causal structu re that captures coordination when simultaneous actions across remote si tes is required. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400014870 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120314T130000 DTEND;TZID="Asia/Jerusalem":20120314T150000 DTSTAMP;TZID="Asia/Jerusalem":20120314T130000 FREEBUSY;FBTYPE=BUSY:20120314T130000/20120314T150000 SUMMARY;LANGUAGE=en-US:The msc talk by David Ben-David about Violation Res olution in Distributed Stream Networks at 2012-03-14 13:00:00 DESCRIPTION;LANGUAGE=en-US:Threshold monitoring applications in distribute d stream networks contin- uously monitor the global score of the network and alert whenever a given threshold is crossed. The network's global sc ore is computed by applying a certain scoring function over the aggregat ed data derived from the network streams. However, the sheer volume and dynamic nature of the streams impose excessive communication overhead. Recently, the concept of local constraints have been presented in which the individual streams are assigned with local constraints that guaranty that as long as all the constraints are valid, the threshold is not cro ssed and thus, no communication over the network is required. While thre shold crossings are fairly rare in most monitoring networks (e.g., natur al hazard monitoring, fire detection, etc.), local constraint violations are not, and in the presence of such, a decision process should take pl ace to determine the network status. This decision process is referred a s violation resolution. Violation resolution should minimize communicati on cost and latency by polling the smallest subset of non-violating stre ams, referred as the resolving set, in the shortest time. We discuss the problem of violation resolution as a minimum resolving set detection problem and suggest bounds for the expected size of the resolving set. In addition, we present a generic algorithm for violation resolution and suggest two instances of it, designed for homogeneous and heterogeneous data setups. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400015000 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120314T143000 DTEND;TZID="Asia/Jerusalem":20120314T163000 DTSTAMP;TZID="Asia/Jerusalem":20120314T143000 FREEBUSY;FBTYPE=BUSY:20120314T143000/20120314T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Ben Livshits SPECIAL GUEST LECTU RE, about Finding Malware on a Web Scale at 2012-03-14 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 539 Taub Bld. UID:111ec4410400014940 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120314T150000 DTEND;TZID="Asia/Jerusalem":20120314T170000 DTSTAMP;TZID="Asia/Jerusalem":20120314T150000 FREEBUSY;FBTYPE=BUSY:20120314T150000/20120314T170000 SUMMARY;LANGUAGE=en-US:The msc talk by Haggai Eran about A Study of Data S tructures with a Deep Heap Shape at 2012-03-14 15:00:00 DESCRIPTION;LANGUAGE=en-US:Computing environments become increasingly para llel, and it seems likely that we will see more cores on tomorrow's deskt ops and server platforms. In a highly parallel system, tracing garbage co llectors may not scale well due to deep heap structures that hinder paral lel tracing. In this work we start by analyzing which data structures mak e current Java benchmarks create deep heap shapes. It turns out that the problem is manifested mostly with benchmarks that employ queues and linke d-lists. We then propose a new construction of the queue data structure w ith extra references that enables better garbage collector parallelism at a low overhead. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400014970 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120314T163000 DTEND;TZID="Asia/Jerusalem":20120314T183000 DTSTAMP;TZID="Asia/Jerusalem":20120314T163000 FREEBUSY;FBTYPE=BUSY:20120314T163000/20120314T183000 SUMMARY;LANGUAGE=en-US:The phd talk by Tamer Salman about Quantum Neural C omputation and Associative Memory at 2012-03-14 16:30:00 DESCRIPTION;LANGUAGE=en-US:This thesis presents quantum analogues of artif icial neural networks. We analyze and compare their performance to know n classical and previously proposed quantum models. First we propose a model for associative memory based on a modification of Grover's quantu m search algorithm and prove that the capacity of the model is exponent ial in the number of bits. We present algorithms for pattern completion and correction and prove that the model does not suffer from spurious mem ories and has a controllable basin of attraction. Then we define mode ls of quantum neurons and devise an algorithm for Hebbian learning through quantum gates. We show that the algorithm implements probabilistic ass ociative memory and probabilistic classification. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400015020 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120319T103000 DTEND;TZID="Asia/Jerusalem":20120319T123000 DTSTAMP;TZID="Asia/Jerusalem":20120319T103000 FREEBUSY;FBTYPE=BUSY:20120319T103000/20120319T123000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Dan Raviv (CS, Tehcnion) abo ut Pixel Club: Geometric constraints in shape analysis at 2012-03-19 10:30 :00 DESCRIPTION;LANGUAGE=en-US:Traditional models of bendable surfaces are bas ed on the exact or approximate invariance to deformations that do not tear or stretch the shape, leaving intact an intrinsic geometry associated wit h it. Intrinsic geometries are typically defined by shortest paths also known as geodesic distances, or diffusion processes on the surface like di ffusion distances. Both methods are implicitly derived from the metric i nduced by the ambient Euclidean space.

Here, we depart from t his restrictive assumption by observing that a different choice of the met ric results in a richer set of geometric invariants. We extend the class ic equi-affine arclength, defined on convex surfaces, to arbitrary shapes with non-vanishing Gaussian curvature. As a result, a family of affine-i nvariant intrinsic geometries is obtained. We propose a computational fr amework that is invariant to the affine group of transformations (similari ty and equi-affine) and thus, by construction, can handle non-rigid sha pes. Diffusion geometry encapsulates the resulting measure to robustly p rovide signatures and computational tools for affine invariant analysis.

The potential of this novel framework is explored in applicati ons such as shape matching and retrieval, symmetry detection, and comput ation of Voroni tessellation.

This work is part of my PhD the sis under the supervision of Ron Kimmel, and was done in collaboration wit h Alexander Bronstein (TAU), Michael Bronstein (USI), Nir Sochen (TAU) a nd Yonatan Aflalo (Technion). ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014880 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120319T153000 DTEND;TZID="Asia/Jerusalem":20120319T173000 DTSTAMP;TZID="Asia/Jerusalem":20120319T153000 FREEBUSY;FBTYPE=BUSY:20120319T153000/20120319T173000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Rick Eads (Agilent Technologies) about ceClub: PCI Express: Technology, Road Map and Design Challenges at 2012-03-19 15:30:00 DESCRIPTION;LANGUAGE=en-US:PCI Express is an industry standard, and the mo st prominent interconnection architecture for I/O devices and other boards inside a computer. It is defined and updated by the PCI-SIG(R) industry s tandards body, originally formed in 1992 as the Peripheral Component Inter connect (PCI) special interest group (SIG). Agilent Technologies, a promin ent developer and provider of test equipment at all levels yet not a compe titor in the computer market, has played a key role in the advancement of PCI Express as a high performance yet extremely reliable interconnection s cheme with very good interoperability among different-vendor equipment.

This talk will provide the following: an overview of PCI Expre ss; PCI Express design and debugging challenges; Agilent tools for PCI Exp ress development (physical layer, signal integrity and protocol validation ); PCI Express electrical and protocol testing; and a live demo of testing solutions. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 861 UID:111ec4410400015090 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120319T183000 DTEND;TZID="Asia/Jerusalem":20120319T203000 DTSTAMP;TZID="Asia/Jerusalem":20120319T183000 FREEBUSY;FBTYPE=BUSY:20120319T183000/20120319T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Victor Kaplansky (I BM R&D, Haifa) about Haifux Club: Software Tools for Reconfigurable Archit ectures at 2012-03-19 18:30:00 DESCRIPTION;LANGUAGE=en-US:Reconfigurable computing allows a lower power c onsumption to achieve higher performance than software, while maintaining a higher level of flexibility than hardware. Reconfigurable devices, such as field-programmable gate arrays (FPGAs), contain an array of computation al elements whose functionality is determined through multiple programmabl e configuration bits. The hardware can be re-programmed to implement an en tirely different circuit.

The focus of the research project ERA (Embedded Reconfigurable Architectures) is to investigate and propose new methodologies in both tools and hardware design to break through the w alls of rising pressure in the demand for performance at the lowest possib le power budget, and help design the next-generation embedded systems plat forms.

In the talk we discuss the proposed strategy to utili ze adaptive hardware to provide the highest possible performance with limi ted power budgets. The envisioned adaptive platform employs a structured d esign approach that allows integration of varying computing elements, netw orking elements, and memory elements. For computing elements, ERA utilizes a mixture of commercially available off-the-shelf processor cores, indust ry-owned IP cores, and application-specific cores. These are dynamically a dapted regarding their composition, organization, and even instruction-set architectures, to provide the best possible performance/power trade-offs. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 6 UID:111ec4410400014900 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120320T113000 DTEND;TZID="Asia/Jerusalem":20120320T133000 DTSTAMP;TZID="Asia/Jerusalem":20120320T113000 FREEBUSY;FBTYPE=BUSY:20120320T113000/20120320T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Ohad Ben-Shahar (Ben-Gurion University) about Pixel Club: Not Your Typical Pixel: On "Cortical Vision " without Visual Cortex at 2012-03-20 11:30:00 DESCRIPTION;LANGUAGE=en-US:Our visual attention is attracted by salient st imuli in our environment and affected by primitive features such as orient ation, color, and motion. Perceptual saliency due to orientation contrast has been extensively demonstrated in behavioral experiments with humans an d other primates and is commonly explained by the very particular function al organization of the primary visual cortex. We challenge this prevailing view by studying orientation-based visual saliency in two non-mammalian s pecies with enormous evolutionary distance to humans. The surprising resul ts not only imply the need to reestablish our understanding of how these p rocesses work at the neural level, but they also suggest that orientation- based saliency has computational optimality in a wide variety of ecologica l contexts, and thus constitutes a universal building block for efficient visual information processing in general.

Disclaimer: While computational vision serves both as inspiration and as an analysis tool in this research, it is far from being the focus of this talk. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400015080 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120321T123000 DTEND;TZID="Asia/Jerusalem":20120321T143000 DTSTAMP;TZID="Asia/Jerusalem":20120321T123000 FREEBUSY;FBTYPE=BUSY:20120321T123000/20120321T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Pavel Hrubes Princeton U niversity) about Theory Seminar: Title: Monotone Unification Problems at 2 012-03-21 12:30:00 DESCRIPTION;LANGUAGE=en-US:Given two monotone polynomials f,g, their unifi er is a pair of monotone polynomials u,v such that f=cu+v and g=u+cv, for some c>0.

The problem I will discuss is: can we have monoton e polynomials f,g which have a unifier, they can be computed by a small mo notone circuit, but every unifier of f,g requires a large monotone circuit ? On one hand, the question is related to arithmetic circuit complexity , and on the other, to the complexity of proofs in subsystems of monotone calculus (or the Frege system). The unification problem can be formulated in several different settings: the simplest is the context of the non-nega tive rank of real matrices. Here, one is required to give quite a strong s eparation between the real rank and the non-negative rank of a matrix - wh ile more modest separations are not known. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400015070 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120321T130000 DTEND;TZID="Asia/Jerusalem":20120321T150000 DTSTAMP;TZID="Asia/Jerusalem":20120321T130000 FREEBUSY;FBTYPE=BUSY:20120321T130000/20120321T150000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Dima Kozakov (Biom edical Engineering, Boston University) about Bioinformatics Forum: Global sampling of Macromolecular Association Energy landscapes using Fast Fourie r Transforms and their Generalizations at 2012-03-21 13:00:00 DESCRIPTION;LANGUAGE=en-US:Sampling energy landscape of macromolecular int eractions is a challenging problem of computational biology. Classical approaches like Molecular Dynamics and Monte Carlo are not always optima l for these applications due to large configurational space of the probl em. On contrary The Fast Fourier Transform (FFT) correlation sampling ap proach can exhaustively evaluate the energies of billions of macromolecu lar configuratins on a grid provided two limitations - the energy is des cribed in the form of a correlation function and the interacting molecul es are assumed rigid. Here we show that by removing above restrictions, i.e adding accurate molecular mechanics energy function and introducing flexibility, FFT based approaches can become method of choice for the mo lecular recognition problems.

We present several applications of the FFT based approaches. First is the problem of protein-protein i nteractions, where our automatic docking server ClusPro was top performi ng server in world wide experiment on blind assesment of protein interac tions (CAPRI). Second is protein hot spot identification method FTMap, w hich is computational analogue of NMR or X-ray based screening using lib raries of fragment-sized compounds. Such experimental approaches include the Multiple Solvent Crystal Structures method used by Ringe and co-wor kers, who soaked protein crystals in the solution of organic solvents, o r the experiments by the Fesik group at Abbott Laboratories, who determi ned protein-ligand interactions by NMR. In both methods, it has been obs erved that the small organic compounds cluster in the active sites of pr oteins.The binding of various compounds provides information on the drug gability of the site, and can be used as input for fragment based drug d esign. Method is shown to agree well with experiments and is applied to binding site identification, finding of druggable sites on protein-pro tein interactions, transmembrane proteins, and nucleic acids. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Seminar room, 4th floor, Emerson building, Technion UID:111ec4410400015170 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120321T143000 DTEND;TZID="Asia/Jerusalem":20120321T163000 DTSTAMP;TZID="Asia/Jerusalem":20120321T143000 FREEBUSY;FBTYPE=BUSY:20120321T143000/20120321T163000 SUMMARY;LANGUAGE=en-US:The msc talk by Hanna Fadida about Automatic Extrac tion of Subcategorization Frames for Hebrew at 2012-03-21 14:30:00 DESCRIPTION;LANGUAGE=en-US:Verb subcategorization frames determine the num ber and types of the syntactic arguments that verbs select, or subcatego rize for. Typically, verbs can be associated with subcategorization frames that specify, for each subcategorized argument, information on the phra ses that can realize this argument. These can be noun phrases (in the c ase of direct objects), prepositional phrases with one or more specific preposition, infinitival verb phrases, clauses introduced by a complementi zer, etc. We introduce a technique and an implemented system for auto matically acquiring Hebrew verb subcategorization frames from textual c orpora. The system generates a large subcategorization lexicon which inc ludes subcategorization frames and frequency information for 3532 verbs. U sing this lexicon was shown to improve two tasks: resolving PP-attachmen t ambiguity, and translating prepositions in an existing statistical mac hine translation system from Arabic to Hebrew. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400015040 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120321T160000 DTEND;TZID="Asia/Jerusalem":20120321T180000 DTSTAMP;TZID="Asia/Jerusalem":20120321T160000 FREEBUSY;FBTYPE=BUSY:20120321T160000/20120321T180000 SUMMARY;LANGUAGE=en-US:The msc talk by Dan Garber about Approximating Semi definite Programs In Sublinear Time at 2012-03-21 16:00:00 DESCRIPTION;LANGUAGE=en-US:Semidefinite programming is a fundamental probl em in convex programming with numerous applications. In the field of combi natorial optimization, many approximation algorithms that rely on sdp have been discovered in the past two decades starting with the work of Goemans and Williamson on MAX-CUT. In the field of machine learning solving sdps is at the heart of many learning tasks such as learning a distance metric and matrix completion. In ML in particular, the amounts of data nowadays a re huge and it is often considered noisy. Thus fast approximation algorith ms are preferable to exact generic solvers, such as interior-point methods , which have impractical running times and memory demands and are not scal able. In this work we present an approximation algorithm for semidefini te programming that produces low-rank solutions with running time that is sublinear in the size of the data. Our method builds on ideas that were de veloped in the theoretical computer science community for approximately so lving Lagrange relaxations of linear programs. We use these ideas with rec ent results in online learning and random sampling to derive our sublinear algorithm. We also present lower bounds that show that the running time o f our algorithm is close to optimal in some cases. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400015030 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120325T163000 DTEND;TZID="Asia/Jerusalem":20120325T183000 DTSTAMP;TZID="Asia/Jerusalem":20120325T163000 FREEBUSY;FBTYPE=BUSY:20120325T163000/20120325T183000 SUMMARY;LANGUAGE=en-US:The CSpecial Talk talk by Shahar Chen (CS, Tehcnion ) about CSpecial Talk: Unified Algorithms for Online Learning and Competit ive Analysis at 2012-03-25 16:30:00 DESCRIPTION;LANGUAGE=en-US:Online learning and competitive analysis are tw o widely studied frameworks for online decision-making settings. Despite t he frequent similarity of the problems they study, there are significan t differences in their assumptions, goals and techniques, hindering a unif ied analysis and richer interplay between the two. In this work, we pro vide several contributions in this direction. We provide a single unified algorithm which by parameter tuning, interpolates between optimal regret for learning from experts (in online learning) and optimal competitive r atio for the metrical task systems problem (MTS) (in competitive analysis) , improving on the results of Blum and Burch (1997). The algorithm also allows us to obtain new regret bounds against ''drifting'' experts, which might be of independent interest. Moreover, our approach allows us to go beyond experts/MTS, obtaining similar unifying results for structured a ction sets and ''combinatorial experts'', whenever the setting has a ce rtain matroid structure.

Joint work with Niv Buchbinder, Seff i Naor, and Ohad Shamir. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400015190 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120326T183000 DTEND;TZID="Asia/Jerusalem":20120326T203000 DTSTAMP;TZID="Asia/Jerusalem":20120326T183000 FREEBUSY;FBTYPE=BUSY:20120326T183000/20120326T203000 SUMMARY;LANGUAGE=en-US:The e-club talk by Ofer Vilenski (Jungo) about e- club: How to Start a Start-Up at 2012-03-26 18:30:00 DESCRIPTION;LANGUAGE=en-US:Have you always dreamed to start a Start-up and didn't know how? Come to meet, hear and ask those who made it big time! ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400015220 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120327T110000 DTEND;TZID="Asia/Jerusalem":20120327T130000 DTSTAMP;TZID="Asia/Jerusalem":20120327T110000 FREEBUSY;FBTYPE=BUSY:20120327T110000/20120327T130000 SUMMARY;LANGUAGE=en-US:The TCE Lecture talk by Tom Leighton (Applied Math, MIT) about TCE Lecture: Consistent Hashing, Danny Lewin, and the Creation of Akamai at 2012-03-27 11:00:00 DESCRIPTION;LANGUAGE=en-US:In 1996, twelve years after making Aliyah to Is rael, Danny Lewin graduated from the Technion and came to MIT to study alg orithms. Over the next few years, he wrote a prize-winning Master’s Thes is on Consistent Hashing and co-founded Akamai Technologies, which today a ccelerates the delivery of over 250,000 web sites, including all of the to p media and commerce brands on the Web.

In this talk, we wil l describe some of Danny’s early work on consistent hashing and how such theoretical work on algorithms ultimately led to the creation of an S&P 5 00 company for accelerating and securing web applications. We will also de scribe some of the scalability and security challenges that the internet i s facing today, and Akamai’s vision for addressing those challenges. < br>
Akamai is headquartered in Cambridge, MA, and through its recent acquisition of Cotendo Networks, Akamai now operates a significant and gr owing R&D center in Tel Aviv.

The talk is dedicated to the me mory of Daniel Lewin, a Technion graduate and entrepreneur, who was killed in the first plane to hit the World Trade Center on 9/11/2001. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Butler Auditorium, Mosad Neeman Building,Technion UID:111ec4410400015060 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120327T143000 DTEND;TZID="Asia/Jerusalem":20120327T163000 DTSTAMP;TZID="Asia/Jerusalem":20120327T143000 FREEBUSY;FBTYPE=BUSY:20120327T143000/20120327T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Prof. Michael Fellows about Para meterizing on the Number of Numbers at 2012-03-27 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400015200 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120328T113000 DTEND;TZID="Asia/Jerusalem":20120328T133000 DTSTAMP;TZID="Asia/Jerusalem":20120328T113000 FREEBUSY;FBTYPE=BUSY:20120328T113000/20120328T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Nadav Amit (CS, Tehcnion) about ceClub: Improving Virtualization Performance Through Phy at 2012-03-28 11: 30:00 DESCRIPTION;LANGUAGE=en-US:Machine virtualization, where many virtual mach ines run on a single physical machine, has become an increasingly popular area of research and development in the last decade. Despite the introduct ion of hardware support for machine virtualization in commodity servers, m any workloads still suffer from degraded performance when run in virtual m achines. This degradation can be particularly acute when an unmodified vir tual machine -- a VM that is not aware it is running in a virtual environm ent -- is used to run the workload. The degradation is is caused primarily by the lack of physical hardware transparency: instead of exposing the un derlying physical hardware transparently to virtual machines, the hypervis or -- the software layer that controls the virtual machines -- usually exp oses hardware "abstractions" instead. This talk will present results from two recent projects that improved VM performance by increasing physical tr ansparency: the first increased device emulation throughput threefold usin g sidecore emulation, and the second improved I/O throughput by 30%-60% us ing direct interrupt delivery to VMs. Joint work with Muli Ben-Yehuda, Abel Gordon, Nadav Har'El, Alex Landau, Assaf Schuster, and Dan Tsafrir. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 401 UID:111ec4410400015260 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120328T123000 DTEND;TZID="Asia/Jerusalem":20120328T143000 DTSTAMP;TZID="Asia/Jerusalem":20120328T123000 FREEBUSY;FBTYPE=BUSY:20120328T123000/20120328T143000 SUMMARY;LANGUAGE=en-US:The phd talk by Gala Yadgar about Cache or Charge? Cooperative Caching with Credits at 2012-03-28 12:30:00 DESCRIPTION;LANGUAGE=en-US:In cooperative storage caching, clients may acc ess blocks directly from each other's caches. Previous studies treated all the cooperating caches as a single pool, maximizing overall system p erformance at the price of possibly degraded performance for individual clients. In light of the popularity of many P2P mechanisms, we re-evalua te the concept of cooperative caching, considering selfish clients that cooperate only if they benefit from the interaction. This is the first stu dy that considers selfish clients in the context of cooperative storage caching. We present and analyze several novel cooperative caching appr oaches for varying degrees of client selfishness. These approaches are b ased on concepts borrowed from distributed storage systems, peer-to-peer systems, and our experience with state-of-the-art multilevel caching me chanisms. Our evaluation focuses on the performance as well as the energ y requirements and operational costs of these approaches, on a wide rang e of systems and two workload domains. Our results provide answers to tw o basic questions: When should clients cooperate, and how? ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400015120 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120328T123000 DTEND;TZID="Asia/Jerusalem":20120328T143000 DTSTAMP;TZID="Asia/Jerusalem":20120328T123000 FREEBUSY;FBTYPE=BUSY:20120328T123000/20120328T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Zohar Karnin (Yahoo! Res earch) about Theory Seminar: On the complexity of the Furthest Hyperplane Problem at 2012-03-28 12:30:00 DESCRIPTION;LANGUAGE=en-US:This paper introduces the Furthest Hyperplane P roblem (FHP), which is an unsupervised counterpart of Support Vector Machi nes. Given a set of n points in R^d, the objective is to produce the hyper plane (passing through the origin) which maximizes the separation margin, that is, the minimal distance between the hyperplane and any input point. To the best of our knowledge, this is the first paper achieving provable r esults regarding FHP. We provide both lower and upper bounds to this NP-ha rd problem. First, we give a simple randomized algorithm whose running tim e is n^{O(1/\theta^2)} where \theta is the optimal separation margin.

We show that its exponential dependency on 1/\theta^2 is tight, up to sub-polynomial factors, assuming SAT cannot be solved in sub-expone ntial time. Next, we give an efficient approximation algorithm. For any \a lpha \in [0,1], the algorithm produces a hyperplane whose distance from at least 1-3\alpha fraction of the points is at least \alpha times the optim al separation margin. Finally, we show that FHP does not admit a PTAS by p resenting a gap preserving reduction from a particular version of the PCP theorem. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400015250 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120329T110000 DTEND;TZID="Asia/Jerusalem":20120329T130000 DTSTAMP;TZID="Asia/Jerusalem":20120329T110000 FREEBUSY;FBTYPE=BUSY:20120329T110000/20120329T130000 SUMMARY;LANGUAGE=en-US:The colloq talk by Judea Pearl, Special Guest Talk, Harvey Prize and Turing Award Winner, about The algorithmization of cause s and counterfactuals at 2012-03-29 11:00:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room Auditorium 2 Taub Bld. UID:111ec4410400015160 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120402T183000 DTEND;TZID="Asia/Jerusalem":20120402T203000 DTSTAMP;TZID="Asia/Jerusalem":20120402T183000 FREEBUSY;FBTYPE=BUSY:20120402T183000/20120402T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Razya Ladelsky (IBM R&D, Haifa) about Haifux Club: Automatic Parallelization in GCC at 2 012-04-02 18:30:00 DESCRIPTION;LANGUAGE=en-US:With the emergence of multicore architectures t here is a growing need for automatic parallelization, that distributes seq uential code into multi threaded code. OpenMP defines language extensions to C, C++, and Fortran for implementing multi-threaded shared memory appli cations. Generation of such extensions by the compiler relieves programmer s from the manual parallelization process. OpenMP specification has been i mplemented in GCC, and is part of the standard release since version 4.2.

In this talk we review the OpenMP and the data dependence su pport which serve as the basic infrastructure for the automatic paralleliz ation in GCC. We describe the capabilities of the automatic parallelizatio n, demonstrated by some examples, and show its benefits with SPEC2006 expe riments. Finally, we discuss current and future directions of work that ma y further extend the optimization's applicability. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 3 UID:111ec4410400014910 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120403T113000 DTEND;TZID="Asia/Jerusalem":20120403T133000 DTSTAMP;TZID="Asia/Jerusalem":20120403T113000 FREEBUSY;FBTYPE=BUSY:20120403T113000/20120403T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Peter Meer (Electrical and C omputer Engineering Department Rutgers University) about Pixel Club: Gen eralized Projection Based M-estimator at 2012-04-03 11:30:00 DESCRIPTION;LANGUAGE=en-US:A new robust estimation algorithm, the generali zed projection based M-estimator (gpbM) is proposed. The algorithm is g eneral and can handle heteroscedastic data where every point in the esti mation have a different covariance. Does not require the user to specif y any (scale) parameters, and can be applied to multiple linear constrai nts for single and multi-carrier problems. The gpbM has three distinct s tages: scale estimation, robust model estimation and inlier/outlier dich otomy. The model estimation stage can be further optimized by using Gras smann manifold theory. For data containing multiple inlier structures th e estimator iteratively determines one structure at a time. We present f our homoscedastic and five heteroscedastic computer vision problems with single or multiple carriers.

Work done together with Sushil Mittal and Saket Anand. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400014920 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120403T133000 DTEND;TZID="Asia/Jerusalem":20120403T153000 DTSTAMP;TZID="Asia/Jerusalem":20120403T133000 FREEBUSY;FBTYPE=BUSY:20120403T133000/20120403T153000 SUMMARY;LANGUAGE=en-US:The Bioinformatics Forum talk by Eugene Kolker (Chi ldren's Hospital, University of Washington) about Bioinformatics Forum: Bi omed Data at Work at 2012-04-03 13:30:00 DESCRIPTION;LANGUAGE=en-US:Access to high-quality data drives our ability to ask questions and find answers. Our lab has a number of projects base d on this premise. Proteomics data from experimental research in diabete s has enabled the discovery of two promising beta cell regulators, Netri n and Sem3a. We have built MOPED, a human and model organisms’ protein expression database, to provide a comprehensive summary of publicly ava ilable proteomics data. With MOPED, users can browse, sort, query, visua lize, and compare their own data against existing studies. Along with ac cess to data comes the need for high-quality analysis of data. We have b uilt SPIRE, a web-based freely available proteomics analysis pipeline to support and expand researchers’ capabilities as they move from data t o knowledge. To further leverage these data resources, the life sciences community has enhanced its collaborations across diverse domains and st rengthened its alliance with computer and data experts. The Data-Enabled L ife Sciences Alliance International (DELSA) was founded to become the le ading voice and coordinating framework to accelerate data-enabled life s ciences research. Finally, the team is also working on improving the man agement and effectiveness of healthcare organizations by mining their ex tensive data resources. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Shelter seminar room, Biology, Technion UID:111ec4410400015280 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120403T143000 DTEND;TZID="Asia/Jerusalem":20120403T163000 DTSTAMP;TZID="Asia/Jerusalem":20120403T143000 FREEBUSY;FBTYPE=BUSY:20120403T143000/20120403T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Professor Frances Rosamond about Determining the Winner of a Dodgson Election is Hard at 2012-04-03 14:30: 00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400015210 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120404T123000 DTEND;TZID="Asia/Jerusalem":20120404T143000 DTSTAMP;TZID="Asia/Jerusalem":20120404T123000 FREEBUSY;FBTYPE=BUSY:20120404T123000/20120404T143000 SUMMARY;LANGUAGE=en-US:The msc talk by Ran Zemach about Algebraic Collocat ion Coarse Approximation (ACCA) in Multigrid at 2012-04-04 12:30:00 DESCRIPTION;LANGUAGE=en-US:Most algebraic multigrid (AMG) methods define t he coarse operators by applying the (Petrov-) Galerkin coarse approximatio n (GCA) where the sparsity pattern and operator complexity of the multigri d hierarchy are dictated by the multigrid transfer operators (prolongation and restriction). Therefore, AMG algorithms must usually settle on some compromise between the quality of the transfer operators and the aggressiv eness of the coarsening, which affect the complexity of the hierarchy of o perators and the overall rate of convergence. A new approach, collocation coarse approximation (CCA), was proposed in 2009 by Wienands and Yavneh, where the coarse approximation is not based on the Galerkin formula and th e choice of the sparsity pattern of the coarse-grid operators is completel y independent of the choice of the transfer operators, P and R. In this work, an algebraic generalization of CCA was studied, leading to a new al gorithm which is fully algebraic and which is based on the aggregation fra mework (smoothed and non-smoothed adaptive aggregation). The algorithm d etermines the coarse-grid operator sparsity pattern using pure aggregation , while it computes the nonzero values using a small set of low-energy eig envectors by a weighted least squares process. Both CCA and ACCA algorith ms are particularly worthwhile for parallel settings. The ACCA algorithm is quite scalable and robust and may be advantageous in cases where strict sparsity constraints prevent us from using high-quality GCA operators. Numerical experiments for two dimensional diffusion problems with sharply varying coefficients, as well as problems with unstructured settings in t wo and in three dimensions, demonstrate the efficacy and potential of this new multigrid algorithm. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400015240 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120404T123000 DTEND;TZID="Asia/Jerusalem":20120404T143000 DTSTAMP;TZID="Asia/Jerusalem":20120404T123000 FREEBUSY;FBTYPE=BUSY:20120404T123000/20120404T143000 SUMMARY;LANGUAGE=en-US:The phd talk by Roy Schwartz about Labelings and Pa rtitions of Graphs at 2012-04-04 12:30:00 DESCRIPTION;LANGUAGE=en-US:The study of combinatorial problems with a subm odular objective function has attracted much attention in recent years, and is partly motivated by the importance of such problems to combinator ial optimization, economics, and algorithmic game theory. A partial list of well-known problems captured by submodular maximization includes: Ma x-Cut, Max-DiCut, Max-k-Cover, Generalized-Assignment, several variants of Max-SAT and some welfare and scheduling problems. While classical wor ks on submodular maximization problems are mostly combinatorial in nature, many recent results are based on continuous algorithmic tools. Usually, the main bottleneck in such a continuous approach is how to approximate ly solve a non-convex relaxation for the submodular problem at hand. We present a new unified continuous greedy algorithm which finds approximat e fractional solutions for both the non-monotone and monotone cases, and improves on the approximation ratio for various applications. Some notabl e immediate implications are information-theoretic tight approximations for Submodular Max-SAT and Submodular-Welfare with k players, for any nu mber of players k, and an improved (1/e)-approximation for maximizing a non-monotone submodular function subject to a matroid or O(1)-knapsack c onstraints. We show that continuous methods can be further used to obtain improved results in other settings. Perhaps the most basic submodular ma ximization problem is the problem of Unconstrained Submodular Maximizati on, which captures some well-studied problems, such as: Max-Cut, Max-DiC ut, and some variants of maximum facility location and Max-SAT. Exploiti ng some symmetry properties of the problem, we present a simple informat ion-theoretic tight (1/2)-approximation algorithm, which unlike previous k nown algorithms keeps a fractional inner state, i.e., it is based on a c ontinuous approach. We note that our algorithm can be further simplified to obtain a purely combinatorial algorithm which runs only in linear ti me. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400015270 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120404T123000 DTEND;TZID="Asia/Jerusalem":20120404T143000 DTSTAMP;TZID="Asia/Jerusalem":20120404T123000 FREEBUSY;FBTYPE=BUSY:20120404T123000/20120404T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Roy Schwartz (CS, Techni on) about Theory Seminar: Labelings and Partitions of Graphs at 2012-04-04 12:30:00 DESCRIPTION;LANGUAGE=en-US:The study of combinatorial problems with a subm odular objective function has attracted much attention in recent years, and is partly motivated by the importance of such problems to combinator ial optimization, economics, and algorithmic game theory. A partial list of well-known problems captured by submodular maximization includes: Ma x-Cut, Max-DiCut, Max-k-Cover, Generalized-Assignment, several variants of Max-SAT and some welfare and scheduling problems.

While cl assical works on submodular maximization problems are mostly combinatorial in nature, many recent results are based on continuous algorithmic tool s. Usually, the main bottleneck in such a continuous approach is how to approximately solve a non-convex relaxation for the submodular problem a t hand. We present a new unified continuous greedy algorithm which finds approximate fractional solutions for both the non-monotone and monotone c ases, and improves on the approximation ratio for various applications. Some notable immediate implications are information-theoretic tight appr oximations for Submodular Max-SAT and Submodular-Welfare with k players, for any number of players k, and an improved (1/e)-approximation for ma ximizing a non-monotone submodular function subject to a matroid or O(1) -knapsack constraints. We show that continuous methods can be further used to obtain improved results in other settings. Perhaps the most basic su bmodular maximization problem is the problem of Unconstrained Submodular Maximization, which captures some well-studied problems, such as: Max-C ut, Max-DiCut, and some variants of maximum facility location and Max-SA T. Exploiting some symmetry properties of the problem, we present a simple information-theoretic tight (1/2)-approximation algorithm, which unlike previous known algorithms keeps a fractional inner state, i.e., it is b ased on a continuous approach. We note that our algorithm can be further simplified to obtain a purely combinatorial algorithm which runs only i n linear time. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400015290 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120404T140000 DTEND;TZID="Asia/Jerusalem":20120404T160000 DTSTAMP;TZID="Asia/Jerusalem":20120404T140000 FREEBUSY;FBTYPE=BUSY:20120404T140000/20120404T160000 SUMMARY;LANGUAGE=en-US:The msc talk by Pavel Bar about Resource Management in Grid Environments at 2012-04-04 14:00:00 DESCRIPTION;LANGUAGE=en-US:Grid computing environments have become mission -critical components in research and industry, offering sophisticated so lutions to exploit large computing and storage resources across multiple geographic locations and administrative domains. Usually, such grid res ources are non-dedicated or opportunistic, as a consequence users will u tilize the resources following a "best effort" approach. However, many r eal-world supercomputing applications, such as computational fluid dynam ics, weather forecasting, and complex system simulations, rely on coallo cation of large numbers of reliable resources as well as on a static and stable execution environment. For such applications the "best effort" q uality of service provided by conventional opportunistic grids is inadeq uate. The research in this area have resulted in the new concept of quas i-opportunistic supercomputing that enables the execution of demanding p arallel applications on a very large number of non-dedicated resources i n grid environments. In this work we propose a complete scheduling fram ework for multi-cluster, heterogeneous environments that provides, in pr actice, an efficient solution for the scheduling and coallocation of top ology-aware applications. The proposed framework is very flexible as it is composed of pluggable components and can be easily configured to supp ort a variety of scheduling policies. We also describe three novel sched uling and coallocation algorithms that were developed and plugged into t he framework. The proposed scheduling framework was integrated into the QosCosGrid system, where it is used as the main decision-making module. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400015230 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120405T113000 DTEND;TZID="Asia/Jerusalem":20120405T133000 DTSTAMP;TZID="Asia/Jerusalem":20120405T113000 FREEBUSY;FBTYPE=BUSY:20120405T113000/20120405T133000 SUMMARY;LANGUAGE=en-US:The phd talk by Yossi Shtok about Adaptive Reconstr uction Methods for Low-Dose Computed Tomography at 2012-04-05 11:30:00 DESCRIPTION;LANGUAGE=en-US:The main problem with contemporary Computed Tom ography (CT) imaging is the high radiation dose absorbed by patients duri ng the screening. Reducing this dose may result in poor quality imaging w hen using the popular fast and direct reconstruction techniques. On the ot her hand, iterative methods powered by statistical models of the scan are better-performing in such cases, but are also very slow. To bridge the g ap between these two solutions, various signal processing techniques that augment the direct reconstruction chain in different ways have been propo sed. In this work we consider a brand of these techniques, which has a le arning capability and involves an off-line example-based training proces s that improves the reconstructed images. Two state-of-the-art noise red uction techniques are adapted to the reconstcruction problem in this mann er. In a different approach, an Artificial Neural Network (ANN) is invoked to build a mixture of experts for CT reconstruction, where the different experts are realized by basic reconstruction methods with varying values of a core parameter controlling their behavior. Our methods show capabil ity of noise and artifacts reduction in low-dose CT images, effectively a llowing for clinical dose savings by factor of ~4. In this supervised l earning setup, we use a direct prior information in the form of high-quali ty reference images which serve as a training set. This is in contrast to almost all existing techniques in CT reconstruction, where at no stage the ground truth is available and the prior information is very implicit and unreliable. An important scenario considered is a Region Of Interest (ROI) reconstruction from limited scan data. Existing techniques offer sub stantial - up to 80% - dose savings, when compared to the standard full-s can reconstruction. We attack this problem using data-adaptive tools and propose an algorithm for local reconstruction that provides a reconstruct ion with excellent locality properties, rendering the more complex techni ques unnecessary. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 337 UID:111ec4410400015180 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120417T143000 DTEND;TZID="Asia/Jerusalem":20120417T163000 DTSTAMP;TZID="Asia/Jerusalem":20120417T143000 FREEBUSY;FBTYPE=BUSY:20120417T143000/20120417T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Oded Lachish about Improved Comp etitive Ratio for the Matroid Secretary Problem at 2012-04-17 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400015360 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120418T113000 DTEND;TZID="Asia/Jerusalem":20120418T133000 DTSTAMP;TZID="Asia/Jerusalem":20120418T113000 FREEBUSY;FBTYPE=BUSY:20120418T113000/20120418T133000 SUMMARY;LANGUAGE=en-US:The ceClub and TCE Lectu talk by Vijay K. Bhargava (University of British Columbia and President, IEEE Communications Society ) about ceClub and TCE Lecture: Green Cellular Networks: A Survey, Some Re search Issues and Challenges at 2012-04-18 11:30:00 DESCRIPTION;LANGUAGE=en-US:In this talk, we present techniques to enable g reen communications in future generation of wireless systems that will rel y on cooperation and cognition to meet increasing demand of high data rate . So far, achieving high data rate has been the primary focus of research in cooperative and CR systems, without much consideration of energy effici ency. However, many of these techniques significantly increase system comp lexity and energy consumption. Escalating energy costs and environmental c oncerns have already created an urgent need for more energy-efficient "gre en" wireless communications. Therefore, we need to design energy-efficient solutions for cooperative and cognitive networks, which will potentially drive the future generation of wireless communication. We focus on several important topics that are crucial towards reducing the energy consumption of the cognitive and cooperative networks. These topics include efficient base station redesign, heterogeneous network deployment, green communicat ions via cognitive radio, cooperative relays to deliver green communicatio ns, and energy efficient cognitive cooperative networks. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400015310 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120419T111500 DTEND;TZID="Asia/Jerusalem":20120419T131500 DTSTAMP;TZID="Asia/Jerusalem":20120419T111500 FREEBUSY;FBTYPE=BUSY:20120419T111500/20120419T131500 SUMMARY;LANGUAGE=en-US:The cggc talk by Tiberiu Popa (Computer Graphics La boratory , ETH, Zurich) about CGGC Seminar: Towards Unencumbered 3D Teleco nferencing Systems at 2012-04-19 11:15:00 DESCRIPTION;LANGUAGE=en-US:With the recent development of auto-multiscopic 3D displays that provide a 3D experience without the use of glasses, and with the recent availability of inexpensive hybrid depth/color cameras such as the Kinect, that provide real-time geometric and texture informati on, we are a step closer to realizing unencumbered 3D teleconferencing systems. However, many challenges still remain to be solved.

In this talk, I will present two teleconferencing software solutions based on the Kinect.

The first is FreeCam - an acquisition system which enables novel-view synthesis in real-time, allowing the viewer to roam a scene using a virtual camera.

The second solution add resses the critical problem in video-mediated teleconferencing of the lack of eye contact caused by the disparity between the locations of the su bject and the camera. While this problem has been partially solved for hig h-end video conferencing systems that employ custom-made expensive hard ware, it has not been convincingly solved for consumer-level systems. In t his talk I will present a real-time gaze correction system that require s only one hybrid depth/color sensor such as the Kinect. This system wi ll be released shortly as a Skype plugin.

Both these projects are collaborations between the Center for Graphics and Geometric Computin g (CGGC) at Technion and the Computer Graphics Lab at ETH Zurich and pa rt of the larger "Being There" Telepresence project administered by the Institute of Media Innovation at Nanyang Technological University in Singa pore. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400015350 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120422T130000 DTEND;TZID="Asia/Jerusalem":20120422T150000 DTSTAMP;TZID="Asia/Jerusalem":20120422T130000 FREEBUSY;FBTYPE=BUSY:20120422T130000/20120422T150000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Benny Applebaum, Tel-Avi v University about Theory Seminar: Pseudorandom Generators with Large Stre tch and Low Locality from Random Local One-Way Functions at 2012-04-22 13: 00:00 DESCRIPTION;LANGUAGE=en-US:Locally-computable pseudorandom generators (PRG s) map n random input bits into m>n pseudorandom bits such that each of th e m outputs depend on a small number of d inputs. While it is known that s uch generators are likely to exist for the case of small sub-linear stretc h m=n+n^{0.9}, it is less clear whether achieving larger stretch is possib le. The existence of such PRGs, which was posed as an open question in pre vious works, has recently gained an additional motivation due to several i nteresting applications in cryptography and complexity. We make progress t owards resolving this question by obtaining several local constructions ba sed on the one-wayness of ``random'' local functions a variant of an assum ption made by Goldreich (ECCC 2000). We will also show that our constructi ons give rise to strong inapproximability results for the densest-subgraph problem in d-uniform hypergraphs for constant d.

The talk d oes not assume prior knowledge of cryptography. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400015340 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120423T140000 DTEND;TZID="Asia/Jerusalem":20120423T160000 DTSTAMP;TZID="Asia/Jerusalem":20120423T140000 FREEBUSY;FBTYPE=BUSY:20120423T140000/20120423T160000 SUMMARY;LANGUAGE=en-US:The msc talk by Shahar Timnat about Concurrent Wait -Free Data-Structrues at 2012-04-23 14:00:00 DESCRIPTION;LANGUAGE=en-US:Lock-free data-structures guarantee overall sys tem progress, whereas wait-free data-structures guarantee the progress o f each and every thread, providing the desirable non-starvation guarante e. While efficient lock-free implementations are known for various dat a- structures, wait-free implementations have been notoriously hard to design and often inefficient. In this talk we first introduce the (firs t) wait-free linked-list, based on Harris's lock-free linked-list algorith m. We then extend this design to work in the fast-path-slow-path method ology in order to a achieive a wait-free linked-list that performs compara bly to the lock-free linked-list. (We might also discuss a generalizatio n of this result and show how to derive effeicient wait-free implementatio ns for many data-structures that lacked such an implementation until recen tly, but I make no such promise!) ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 701 UID:111ec4410400015380 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120423T183000 DTEND;TZID="Asia/Jerusalem":20120423T203000 DTSTAMP;TZID="Asia/Jerusalem":20120423T183000 FREEBUSY;FBTYPE=BUSY:20120423T183000/20120423T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Gabor Szabo about H aifux Club: Modern Web Development in Perl at 2012-04-23 18:30:00 DESCRIPTION;LANGUAGE=en-US:Back in the early years of the dynamic web, Per l was the de-facto
choice of language.

It was so ubique th at people associated it with CGI. For that, even today, Perl
suffers from a bad image. Even though, under the visible surface,
there are strong powers leading to a much better future.

In the last c ouple of years a new bread of frameworks appeared in the
Perl world.

They are all based on the PSGI standard that rhymes to WSGI and Rack.

In this presentation we will see a couple of exampl es writing
PSGI  (http://plackperl.org/) and
Dancer (http://perldancer.org/) based applications. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 3 UID:111ec4410400015320 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120424T143000 DTEND;TZID="Asia/Jerusalem":20120424T163000 DTSTAMP;TZID="Asia/Jerusalem":20120424T143000 FREEBUSY;FBTYPE=BUSY:20120424T143000/20120424T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Eugene Agichtein about Extractin g Meaning from Search Behavior Data at 2012-04-24 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400015330 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120501T113000 DTEND;TZID="Asia/Jerusalem":20120501T133000 DTSTAMP;TZID="Asia/Jerusalem":20120501T113000 FREEBUSY;FBTYPE=BUSY:20120501T113000/20120501T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Alex Bronstein (School of El ectrical Engineering, Tel Aviv University) about Pixel Club: Learning Fea ture Descriptors at 2012-05-01 11:30:00 DESCRIPTION;LANGUAGE=en-US:SIFT-like local feature descriptors are ubiquit ously employed in such computer vision applications as content-based retri eval, video analysis, copy detection, object recognition, photo-tourism, a nd 3D reconstruction from multiple views. Feature descriptors can be desig ned to be invariant to certain classes of photometric and geometric transf ormations, in particular, affine and intensity scale transformations. Ho wever, real transformations that an image can undergo can only be approxim ately modeled in this way, and thus most descriptors are only approximatel y invariant in practice. Secondly, descriptors are usually high-dimensiona l (e.g. SIFT is represented as a 128-dimensional vector). In large-scale r etrieval and matching problems, this can pose challenges in storing and re trieving descriptor data. In this talk, we will show how to map the descri ptor vectors into the Hamming space, in which the Hamming metric is used t o compare the resulting representations. This way, we reduce the size of t he descriptors by representing them as short binary strings and learn desc riptor invariance from examples. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400015300 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120501T143000 DTEND;TZID="Asia/Jerusalem":20120501T163000 DTSTAMP;TZID="Asia/Jerusalem":20120501T143000 FREEBUSY;FBTYPE=BUSY:20120501T143000/20120501T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Yossi Gil about How Do Classes C hoose Their Parents? (Preferential Attachment to the Trial) at 2012-05-01 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400015400 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120502T123000 DTEND;TZID="Asia/Jerusalem":20120502T143000 DTSTAMP;TZID="Asia/Jerusalem":20120502T123000 FREEBUSY;FBTYPE=BUSY:20120502T123000/20120502T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Ankit Gupta (CS, Technio n) about Theory Seminar: Random Arithmetic Formulas can be Reconstructed Efficiently at 2012-05-02 12:30:00 DESCRIPTION;LANGUAGE=en-US:What is an optimal formula computing a given mu ltivariate polynomial $f$? In this work, we show that this question admits an efficient algorithmic solution in an average-case sense.

Specifically, we consider the following situation. Let $\F$ be a field, $ S \subseteq \F$ be a finite subset of field elements, $\vecX =(X_1, X_2, \ ldots, X_n) $ be a tuple of formal variables and $\Delta \geq 0 $ be an in teger representing the product-depth of the hidden (unknown) formula. In t his work we define "a random formula" as one generated by the following pr ocess. If $\Delta=0$ then pick $(n+1)$ field elements $a_0, a_1, \ldots, a _n$ uniformly at random from $S$ and output a node labelled with the affin e form $(a_0 + a_1 X_1 + a_2 X_2 + ... + a_n X_n)$. If $\Delta > 0$ then recursively pick four random subformulas $A,B,C$ and $D$ of product depth $(\Delta-1)$ each and output $(A \cdot B + C \cdot D)$. Thus it’s a bina ry formula which is layered with alternating layers of addition and multip lication gates and where the leaf nodes are labelled with affine forms ove r the variable set. We present a randomized algorithm that, given blackbox access to just the output polynomial $f$ of such a formula $\phi$, can ef ficiently reconstruct back (an equivalent) formula of the same size in tim e $\poly(n \cdot \size(\phi))$ and with the probability of success (over t he random choice of $\phi$) being at least $\left( 1 - \frac{d^{O(1)}}{\ab s{S}} \right)$.

This, then, is the strongest model of arithme tic computation for which a reconstruction algorithm is presently known, a lbeit efficient in a distributional sense rather than in the worst case. T he main technical work involved in devising the above algorithm is underst anding and characterizing the high-dimensional components of the singular locus of an arbitrary linear combination of (a few) random formulas.

(joint work with Neeraj Kayal, Youming Qiao) ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400015390 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120507T183000 DTEND;TZID="Asia/Jerusalem":20120507T203000 DTSTAMP;TZID="Asia/Jerusalem":20120507T183000 FREEBUSY;FBTYPE=BUSY:20120507T183000/20120507T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Uri Barkan about Ha ifux Club: Scientific Python at 2012-05-07 18:30:00 DESCRIPTION;LANGUAGE=en-US:See Hebrew announcement. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 3 UID:111ec4410400015420 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120508T113000 DTEND;TZID="Asia/Jerusalem":20120508T133000 DTSTAMP;TZID="Asia/Jerusalem":20120508T113000 FREEBUSY;FBTYPE=BUSY:20120508T113000/20120508T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Daniel Glazner (Faculty of M athematics and Computer Science The Weizmann Institute of Science) about Pixel Club: Viewpoint-Aware Object Detection and Pose Estimation at 2012- 05-08 11:30:00 DESCRIPTION;LANGUAGE=en-US:We describe an approach to category-level detec tion and viewpoint estimation for rigid 3D objects from single 2D images . In contrast to many existing methods, we directly integrate 3D reasoni ng with an appearance-based voting architecture. Our method relies on a nonparametric representation of a joint distribution of shape and appear ance of the object class. Our voting method employs a novel parametrizat ion of joint detection and viewpoint hypothesis space, allowing efficien t accumulation of evidence. We combine this with a re-scoring and refine ment mechanism, using an ensemble of view-specific Support Vector Machin es. We evaluate the performance of our approach in detection and pose es timation of cars on a number of benchmark datasets.

This is j oint work with Meirav Galun, Sharon Alpert, Ronen Basri and Gregory Shakhn arovich. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400015450 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120508T143000 DTEND;TZID="Asia/Jerusalem":20120508T163000 DTSTAMP;TZID="Asia/Jerusalem":20120508T143000 FREEBUSY;FBTYPE=BUSY:20120508T143000/20120508T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Guy Kindler about Functions whic h are close to being k-local are juntas at 2012-05-08 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400015430 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120509T113000 DTEND;TZID="Asia/Jerusalem":20120509T133000 DTSTAMP;TZID="Asia/Jerusalem":20120509T113000 FREEBUSY;FBTYPE=BUSY:20120509T113000/20120509T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Chen Avin (Ben Gurion University ) about ceClub: Self-Adjusting Networks and Distributed Data Structures at 2012-05-09 11:30:00 DESCRIPTION;LANGUAGE=en-US:We study self-optimizing networks and data stru ctures. The goal of this research line is the design of fully distributed algorithms which flexibly adapt the network to a dynamic environment such as changing demands or traffic patterns. The idea of self-adjusting networ ks is motivated by trends in today's Internet like large data centers and peer-to-peer networks.

In this talk I will preset the genera l model of the problem and some initial theoretical results on data center s placement and splay networks and the connection to well known concepts s uch as minimum linear arrangement, splay trees and Entropy.

Joint work with: Michael Borokhovich (BGU), Bernhard Haeupler (MIT), Zvi L otker (BGU), Christian Scheideler (U. Paderborn) and Stefan Schmid (T-Labs ). ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 861 UID:111ec4410400015490 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120509T123000 DTEND;TZID="Asia/Jerusalem":20120509T143000 DTSTAMP;TZID="Asia/Jerusalem":20120509T123000 FREEBUSY;FBTYPE=BUSY:20120509T123000/20120509T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Shiri Chechik (Weizmann Institute of Science) about Theory Seminar: Fully Dynamic Approximate Dist ance Oracles for Planar Graphs via Forbidden-Set Distance Labels at 2012-0 5-09 12:30:00 DESCRIPTION;LANGUAGE=en-US:Distance oracle is a data structure that provid es fast answers to distance queries. Recently, the problem of designing distance oracles capable of answering restricted distance queries, that i s, estimating distances on a subgraph avoiding some forbidden vertices, ha s attracted a lot of attention. In this talk, we will consider forbidden s et distance oracles for planar graphs. I’ll present an efficient compact distance oracle that is capable of handing any number of failures.

In addition, we will consider a closely related notion of fully dynamic distance oracles. In the dynamic distance oracle problem ins tead of getting the failures in the query phase, we rather need to handle an adversarial online sequence of update and query operations. Each query operation involves two vertices s and t whose distance needs to be estimat ed. Each update operation involves inserting/deleting a vertex/edge from t he graph.

I’ll show that our forbidden set distanc e oracle can be tweaked to give fully dynamic distance oracle with improve d bounds compared to the previously known fully dynamic distance oracle fo r planar graphs.

Joint work with Ittai Abraham and C yril Gavoille. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400015470 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120514T183000 DTEND;TZID="Asia/Jerusalem":20120514T203000 DTSTAMP;TZID="Asia/Jerusalem":20120514T183000 FREEBUSY;FBTYPE=BUSY:20120514T183000/20120514T203000 SUMMARY;LANGUAGE=en-US:The Haifux, Linux Haifa talk by Dotan Barak (Mellan ox) about Haifux Club: InfiniBand, RoCE and RDMA Verbs - Empowering Superc omputing and Data Center Interconnects at 2012-05-14 18:30:00 DESCRIPTION;LANGUAGE=en-US:This lecture aims to provide a brief introducti on to the InfiniBand™ architecture and programming using RDMA verbs. I nfiniBand is an open standard, used in HPC (supercomputing) and data cen ter environments for high performance connectivity.

This standard defines a complete fabric architecture, from the physical laye r all the way to the programming API. The programming API, also known as RDMA verbs, allows for transparent memory operations over the network, transport layer offloading, and a complete kernel bypass.

I n this lecture we will also describe the intricate relationship between InfiniBand and open source projects. We will review the basics of the RD MA verbs, correlate the InfiniBand world to the Ethernet-TCP/IP-sockets world, and show a full data flow over this technology. Finally, we will show how RDMA verbs are also available for Ethernet networks through a s tandard called RoCE.

This standard defines a complete fabric architecture, from the physical layer all the way to the programming API . The programming API, also known as RDMA verbs, allows for transparent memory operations over the network, transport layer offloading, and a complete kernel bypass. In this lecture we will also describe the int ricate relationship between InfiniBand and open source projects. We will review the basics of the RDMA verbs, correlate the InfiniBand world to the Ethernet-TCP/IP-sockets world, and show a full data flow over this technology. Finally, we will show how RDMA verbs are also available for Ethernet networks through a standard called RoCE. This standard defi nes a complete fabric architecture, from the physical layer all the way to the programming API. The programming API, also known as RDMA verbs, a llows for transparent memory operations over the network, transport laye r offloading, and a complete kernel bypass. In this lecture we will also describe the intricate relationship between InfiniBand and open so urce projects. We will review the basics of the RDMA verbs, correlate th e InfiniBand world to the Ethernet-TCP/IP-sockets world, and show a full data flow over this technology. Finally, we will show how RDMA verbs ar e also available for Ethernet networks through a standard called RoCE. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 3 UID:111ec4410400015480 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120516T113000 DTEND;TZID="Asia/Jerusalem":20120516T133000 DTSTAMP;TZID="Asia/Jerusalem":20120516T113000 FREEBUSY;FBTYPE=BUSY:20120516T113000/20120516T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Alex Shraer (Yahoo! Research) a bout ceClub:Dynamic Reconfiguration of Primary/Backup Clusters at 2012-05- 16 11:30:00 DESCRIPTION;LANGUAGE=en-US:Dynamically changing (reconfiguring) the member ship of a replicated distributed system while preserving data consistency and system availability is a challenging problem. In this talk I will disc uss this problem in the context of Primary/Backup clusters and Apache Zook eeper. Zookeeper is an open source system which enables highly reliable di stributed coordination. It is widely used in industry, for example in Yaho o!, Facebook,Twitter, VMWare, Box, Cloudera, Mapr, UBS, Goldman Sachs, Nic ira, Netflix and many others. A common use-case of Zookeeper is to dynamic ally maintain membership and other configuration metadata for its users. Z ookeeper itself is a replicated distributed system. Unfortunately, the mem bership and all other configuration parameters of Zookeeper are static - t hey're loaded during boot and cannot be altered. Operators resort to ''rol ling restart'' - a manually intensive and error-prone method of changing t he configuration that has caused data loss and inconsistency in production . Automatic reconfiguration functionality has been requested by operators since 2008. Several previous proposals were found incorrect and rejected. We designed and implemented a new reconfiguration protocol in Zookeeper an d are currently integrating it into the codebase. It fully automates confi guration changes: the set of Zookeeper servers, their roles, addresses, et c. can be changed dynamically, without service interruption and while main taining data consistency. By leveraging the properties already provided by Zookeeper our protocol is considerably simpler than state of the art in r econfiguration protocols. Our protocol also encompasses the clients -- cli ents are rebalanced across servers in the new configuration, while keeping the extent of migration proportional to the change in membership.

This is a joint work with Benjamin Reed (Yahoo!), Dahlia Malkhi (MS R) and Flavio Junqueira (Yahoo!). A paper describing this work will appear in the 2012 Usenix ATC conference and in the 2012 Hadoop Summit. http://w ww.cs.technion.ac.il/~shralex/zkreconfig.pdf ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400015500 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120516T123000 DTEND;TZID="Asia/Jerusalem":20120516T143000 DTSTAMP;TZID="Asia/Jerusalem":20120516T123000 FREEBUSY;FBTYPE=BUSY:20120516T123000/20120516T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Ariel Gabizon (CS. Techn ion) about Theory Seminar: Extractors for Polynomials Sources over Consta nt-Size Fields of Small Characteristic at 2012-05-16 12:30:00 DESCRIPTION;LANGUAGE=en-US:Let F be the field of q elements, where q=p^l f or prime p. Informally speaking, a polynomial source is a distribution o ver F^n sampled by low degree multivariate polynomials. In this paper, w e construct extractors for polynomial sources over fields of constant size q assuming p<
For instance, suppose a distribution X over F^n has support size q^k and is sampled by polynomials of individua l degree d and total degree D. When p, D and the "entropy rate" k/n are constant, we get an extractor over constant-size fields with constant e rror. The only previous construction by Dvir, Gabizon and Wigderson requ ired a field of size polynomial in n.

Our proof follows sim ilar lines to that of DeVos and Gabizon on extractors for affine sources, i.e., polynomial sources of degree 1. Our result makes crucial use of a t heorem of Hou, Leung and Xiang giving a lower bound on the dimension of pr oducts of subspaces. The key insights that enable us to extend these resul ts to the case of polynomial sources of degree greater than 1 are
1) A source with support size q^k must have a linear span of dimensio n at least k, and in the setting of low-degree polynomial sources it s uffices to increase the dimension of this linear span.

2)Dist inct Frobenius automorphisms of a (single) low-degree polynomial source are `pseudo-independent' in the following sense: Taking the product of d istinct automorphisms (of the very same source) increases the dimension of the linear span of the source.

Joint work with Eli Ben-Sas son ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400015510 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120522T113000 DTEND;TZID="Asia/Jerusalem":20120522T133000 DTSTAMP;TZID="Asia/Jerusalem":20120522T113000 FREEBUSY;FBTYPE=BUSY:20120522T113000/20120522T133000 SUMMARY;LANGUAGE=en-US:The pixel-club talk by Micha Livne (CS, University of Toronto) about Pixel Club: Human Attributes from 3D Pose Tracking at 20 12-05-22 11:30:00 DESCRIPTION;LANGUAGE=en-US:This talk concerns the estimation of human attr ibutes from 3D human pose and motion. We consider both physical attribut es (eg, gender and weight) and aspects of mental state (eg, mood). This task is useful for man-machine communication, and it provides a natural benchmark for evaluating the performance of 3D pose tracking methods. Ba sed on an extensive corpus of motion capture data, with physical and per ceptual ground truth, we analyze the inference of subtle biologically-in spired attributes from cyclic gait data. It is shown that inference is a lso possible with partial observations of the body, and with motions as short as a single gait cycle. Learning models from small amounts of nois y video pose data is, however, prone to over-fitting. To mitigate this w e formulate learning in terms of domain adaptation, for which mocap data is uses to regularize models for inference from video-based data. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1061 UID:111ec4410400015520 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120522T143000 DTEND;TZID="Asia/Jerusalem":20120522T163000 DTSTAMP;TZID="Asia/Jerusalem":20120522T143000 FREEBUSY;FBTYPE=BUSY:20120522T143000/20120522T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Yuval Ishai about The Complexity of Cryptography at 2012-05-22 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400015440 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120523T113000 DTEND;TZID="Asia/Jerusalem":20120523T133000 DTSTAMP;TZID="Asia/Jerusalem":20120523T113000 FREEBUSY;FBTYPE=BUSY:20120523T113000/20120523T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Reuven Cohen (CS, Technion) abou t ceClub: Dr. Philip M. Merlin Memorial Lecture and Prize Award: Schedulin g Algorithms for the 4G Cellular Networks at 2012-05-23 11:30:00 DESCRIPTION;LANGUAGE=en-US:Cellular networks are becoming more and more cr ucial to our daily lives and operators are seeking new technologies for in creasing their bandwidth. Examples for such technologies are cell sectoriz ation, fractional frequency reuse, and coordinated multipoint Tx/Rx. To ta ke advantage of these new technologies, the scheduler logic at the base st ation needs to determine not only when to transmit each packet but also wh at modulation and coding scheme to use, which frequency reuse area's resou rces, and by which transmitting antenna. In this talk, I will discuss all these modern scheduling problems and present algorithms for solving them ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:EE Meyer Building 1003 UID:111ec4410400015530 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120523T123000 DTEND;TZID="Asia/Jerusalem":20120523T143000 DTSTAMP;TZID="Asia/Jerusalem":20120523T123000 FREEBUSY;FBTYPE=BUSY:20120523T123000/20120523T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Elad Haramaty ( (CS, Tec hnion) about Theory Seminar: Optimal Testing of Multivariate Polynomials over Small Prime Field at 2012-05-23 12:30:00 DESCRIPTION;LANGUAGE=en-US:We consider the problem of testing if a given f unction f is close to a n-variate degree d polynomial over the finite fie ld of q elements. The natural, low-query, test for this property would be to pick the smallest dimension t= t(q,d )≈ d/q such that every function of degree greater than d reveals this feature on some t-dimensional affine subspace and to test that f when restricted to a random t-dimensional aff ine subspace is a polynomial of degree at most d on this subspace. Such a test makes only qt queries, independent of n. Previous works, by Alon et a l., and Kaufman and Ron and Jutla et al. , showed that this natural test r ejected functions that were Ω(1)-far from degree d-polynomials with proba bility at least Ω(q−t) (the results of hold for all fields, while the results of hold only for fields of prime order). Thus to get a constant pr obability of detecting functions that were at constant distance from the s pace of degree d polynomials, the tests made q^{2t} queries. Kaufman and R on also noted that when q is prime, then q^t queries are necessary. Thus t hese tests were off by at least a quadratic factor from known lower bounds . It was unclear if the soundness analysis of these tests were tight and t his question relates closely to the task of understanding the behavior of the Gowers Norm. This motivated the work of Bhattacharyya et al., who gave an optimal analysis for the case of the binary field and showed that the natural test actually rejects functions that were Ω(1)-far from degree d- polynomials with probability at least Ω(1).

In this work we give an optimal analysis of this test for all fields showing that the natu ral test does indeed reject functions that are Ω(1)-far from degree d pol ynomials with Ω(1)-probability. Our analysis thus shows that this test is optimal (matches known lower bounds) when q is prime. Our approach extend s the proof technique of Bhattacharyya et al., however it has to overcome many technical barriers in the process. The natural extension of their ana lysis leads to an O(q^d) query complexity, which is worse than that of Kau fman and Ron for all q except 2! The main technical ingredient in our work is a tight analysis of the number of “hyperplanes” (affine subspaces of co-dimension 1) on which the restriction of a degree dpolynomial has de gree less than d. We show that the number of such hyperplanes is at most O (qtq,d ) — which is tight to within constant factors.

Joint work with Amir Shpilka and Madhu Sudan. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400015540 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120529T143000 DTEND;TZID="Asia/Jerusalem":20120529T163000 DTSTAMP;TZID="Asia/Jerusalem":20120529T143000 FREEBUSY;FBTYPE=BUSY:20120529T143000/20120529T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Zvi Galil about Computing in the 21st Century: The Georgia Tech Model at 2012-05-29 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400015410 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120530T123000 DTEND;TZID="Asia/Jerusalem":20120530T143000 DTSTAMP;TZID="Asia/Jerusalem":20120530T123000 FREEBUSY;FBTYPE=BUSY:20120530T123000/20120530T143000 SUMMARY;LANGUAGE=en-US:The Theory Seminar talk by Ofer Neiman (Ben-Gurion University) about Theory Seminar: From Irreducible Representations to Loca lly Decodable Codes at 2012-05-30 12:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 201 UID:111ec4410400015590 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120530T133000 DTEND;TZID="Asia/Jerusalem":20120530T153000 DTSTAMP;TZID="Asia/Jerusalem":20120530T133000 FREEBUSY;FBTYPE=BUSY:20120530T133000/20120530T153000 SUMMARY;LANGUAGE=en-US:The msc talk by Haggai Toledano about Coverage-Driv en Refinement of Conceptual Representations at 2012-05-30 13:30:00 DESCRIPTION;LANGUAGE=en-US:Many text processing tasks are based on estimat ing semantic relatedness between texts. For example, in information retrie val, relevancy of documents can be determined based on the semantic distan ce from the query. Recently, many algorithms have been developed for evalu ating semantic relatedness based on a conceptual representation of the inp ut texts. The concept spaces for these algorithms are based, in most cases , on large repositories of knowledge, such as Wikipedia and Wordnet. These large concept spaces often yield representations that consist of very lar ge collections of concepts, that in many cases have a negative impact on t he performance of the semantic tasks due to redundancy that gives a superf icially large weight to less relevant concepts, thus hiding important sema ntic aspects of the texts. In this work we present a new algorithm for con ceptual representations that are based on hierarchical concept spaces. The algorithm incrementally adds strongly-associated concepts to the represen tation, while using the hierarchical structure of the semantic database to maximize coverage. We test the new algorithm for text relatedness tasks a nd show its advantage over existing approaches. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400015560 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120530T163000 DTEND;TZID="Asia/Jerusalem":20120530T183000 DTSTAMP;TZID="Asia/Jerusalem":20120530T163000 FREEBUSY;FBTYPE=BUSY:20120530T163000/20120530T183000 SUMMARY;LANGUAGE=en-US:The msc talk by Amir Abboud about Safe Zones: An Ef ficient Approach to Distributed Monitoring at 2012-05-30 16:30:00 DESCRIPTION;LANGUAGE=en-US:Many monitoring tasks over distributed data str eams can be formulated as a continuous query using a function that is d efined over the global average of data vectors derived from the streams. The query will typically produce an alert when the value of the func tion crosses a predefined threshold. A fundamental problem in efficient scalable implementation of such threshold queries is that the data stre ams are distributed, sometimes over a wide geographical region. Moving all the data to a centralized data center for query processing may inc ur infeasible communication overheads and inflated data center resource costs. In some cases it may be prohibited altogether by the sheer aggr egated size of the data, or by privacy laws. The goal is thus to enha nce scalability by processing the query locally, using as little communi cation and global coordination as possible. We present a novel sch eme for communication reduction in distributed monitoring using local co nstraints. Communication and global coordination are required only in t he event that the local constraints are violated by the incoming data. Our work improves on previous work in a few critical aspects. First, whereas previous work required constructing a ``distributed cover'' of t he entire convex hull of the local data vectors, our work compiles const raints that are designed to cover only the global average; further, they are directly matched and tailored to fit the local data distribution a t each stream. The result is a dramatic decrease in the required volume of communication compared to previous state of the art, up to two orders of magnitude in our experiments with real-life data. Both the experime nts and theoretical study suggest that the improvement factor increases with the dimension of the data. Also, in contrast to previous work, wh ich necessitated complicated constraints and required enormous computati onal effort over each of the streams, our scheme can use very simple con straints which incur negligible local overhead. This latter advantage makes our new approach applicable to thin, battery-operated sensors and cellular devices. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Taub 601 UID:111ec4410400015550 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120603T130000 DTEND;TZID="Asia/Jerusalem":20120603T150000 DTSTAMP;TZID="Asia/Jerusalem":20120603T130000 FREEBUSY;FBTYPE=BUSY:20120603T130000/20120603T150000 SUMMARY;LANGUAGE=en-US:The cggc talk by Daniel Reem (IMPA, Rio, Brasil) ab out CGGC Seminar: On the possibility of simple parallel computing of Voron oi diagrams and Delaunay Graphs at 2012-06-03 13:00:00 DESCRIPTION;LANGUAGE=en-US:Although many algorithms for computing Euclidea n Voronoi diagrams of point sites have been published, most of them are se quential in nature and hence cast inherent difficulties on the possibility to compute the diagrams in parallel. We present a new algorithm which ena bles the (combinatorial) computation of each of the Voronoi cells independ ently of the other ones. The algorithm is significantly different from pre vious ones and some of the ideas related to it are in the spirit of convex analysis.

A new combinatorial structure for representing the ce lls is described along the way, and the computation of the corresponding D elaunay graph follows as a simple consequence.

An implementati on of the algorithm (done by Omri Azencot) will be presented. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400015370 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120605T143000 DTEND;TZID="Asia/Jerusalem":20120605T163000 DTSTAMP;TZID="Asia/Jerusalem":20120605T143000 FREEBUSY;FBTYPE=BUSY:20120605T143000/20120605T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Micha Hofri about The surprises of Quick Sort at 2012-06-05 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400015570 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120613T113000 DTEND;TZID="Asia/Jerusalem":20120613T133000 DTSTAMP;TZID="Asia/Jerusalem":20120613T113000 FREEBUSY;FBTYPE=BUSY:20120613T113000/20120613T133000 SUMMARY;LANGUAGE=en-US:The ceClub talk by Christoph Lenzen (Weizmann Insti tue of Science) about ceClub: Improved Bounds for Byzantine Self-Stabilizi ng Clock Synchronization at 2012-06-13 11:30:00 DESCRIPTION;LANGUAGE=en-US:The challenging task of Byzantine self-stabiliz ing pulse synchronization requires that, in the presence of a minority of nodes that are permanently maliciously faulty, the non-faulty nodes must s tart firing synchronized pulses in a regular fashion after a finite amount of time, regardless of the initial state of the system. We study this pro blem under full connectivity in a model where nodes have local clocks of u nknown, but bounded drift, and messages are delayed for an unknown, but bo unded time.

We present a generic scheme that, given a synchr onous consensus protocol P, defines a self-stabilizing pulse synchronizati on algorithm A(P). If P terminates within R rounds (deterministically or w ith high probability), A(P) stabilizes within O(R) time (deterministically or with high probability, respectively). Utilizing different consensus ro utines, our transformation yields various improvements over previous techn iques in terms of stabilization time and bit complexity. Finally, we sketc h how to establish the abstraction of synchronous, integer-labeled rounds on top of pulse synchronization, at essentially the same complexity bounds .

We will discuss our approach and its merits assuming no pr evious knowledge on the problem, however, basic familiarity with consensus will be beneficial. ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400015600 END:VEVENT BEGIN:VEVENT DTSTART;TZID="Asia/Jerusalem":20120619T143000 DTEND;TZID="Asia/Jerusalem":20120619T163000 DTSTAMP;TZID="Asia/Jerusalem":20120619T143000 FREEBUSY;FBTYPE=BUSY:20120619T143000/20120619T163000 SUMMARY;LANGUAGE=en-US:The colloq talk by Rina Dechter about Principles of Reasoning with Graphical Models at 2012-06-19 14:30:00 DESCRIPTION;LANGUAGE=en-US: ATTENDEE;CUTYPE=GROUP;PARTSTAT=TENTATIVE:mailto:webmaster@cs.technion.ac.il LOCATION:Room 337-8 Taub Bld. UID:111ec4410400015460 END:VEVENT END:VCALENDAR