The Taub Faculty of Computer Science Events and Talks

Hierarchical modeling is often applied in fields such as computer vision, computer graphics, computer-aided geometric design, cartography and virtual reality. The importance of this approach has increased in recent years due to the ability to create models with more details. Hierarchical modeling is useful in order to make storage, transmission, computation, and visualization of these models possible and more efficient. The most prevalent 3D hierarchical geometric data structure is the Octree. In this research, the ...
[Full version]

In recent years, several works have demonstrated how the study of biology can benefit from an algorithmic perspective. In this talk I discuss a new approach for such a methodology based on combining theoretical tradeoff techniques with field experiments to obtain bounds on biological parameters. A proof of concept for this framework is provided by considering central search foraging strategies of desert ants, and obtaining theoretical tradeoffs between the search time and the memory complexity ...
[Full version]

The current trend to move from homogeneous to heterogeneous multi-core (HetCMP) systems promises further performance and energy-efficiency benefits. A typical HetCMP system includes two distinct types of cores, such as high performance sophisticated ("large") cores and simple low-power ("small") cores. In those heterogeneous platforms, execution phases of application threads that are CPU-intensive can take best advantage of large cores, whereas I/O or memory intensive execution phases are best suited and assigned to small cores. However, ...
[Full version]

Compressive sensing (CS), also known as compressive sampling, has become widely popular in recent years. In the first part of the talk, we review the little known fact, that the invention of CS preceded the papers that popularized it by almost a decade. Spectrum-blind sampling (SBS), proposed by Bresler and Feng in the mid-90’s, and further developed into “image compression on the fly,” with Venkataramani, and Gastpar, is the first known compressed sensing technique. This ...
[Full version]

We consider the problem of finding a few representatives for a dataset, i.e., a subset of data points that efficiently describes the entire dataset. We assume that each data point can be expressed as a linear combination of the representatives and formulate the problem of finding the representatives as a sparse multiple measurement vector problem. In our formulation, both the dictionary and the measurements are given by the data matrix, and the unknown sparse codes ...
[Full version]

The open source production process is fascinating in many respects: it produces high quality code outside the boundaries of the hierarchical organization, it defies the rules of many traditional software production models and it has many profound implications on social, political, and economic systems of the Internet era. I will discuss prior research work on the topic. For example, should open source serve as an example of how to develop systems of participation with the ...
[Full version]

Digital maps and devices with an integrated GPS receiver, such as smartphones, are now an integral part of our lives. We deal with two practical problems related to GPS trajectories and digital maps. The first problem is the classical problem of map-matching; namely, matching a given GPS trajectory, possibly noisy or sparse, to the sequence of actual roads traversed by the carrier of the GPS receiver. We provide a novel solution to this problem. Our ...
[Full version]

The advent of large-scale wireless networks has presented the distributed computing community with a new challenge: wireless networks are much more disorderly than traditional wired networks, and they are difficult to model and predict. In addition, they are subject to different design considerations: among other criteria, it is crucial to conserve power by reducing the amount of communication between network nodes. This communication constraint leads to interesting connections between distributed computing and communication complexity. In ...
[Full version]

Extremely slow, or straggler, tasks are a major performance bottleneck in map-reduce systems. Hadoop infrastructure makes an effort to both avoid them (through minimizing remote data accesses) and handle them in the runtime (through speculative execution). However, the mechanisms in place neither guarantee the avoidance of performance hotspots in task scheduling, nor provide any easy way to tune the timely detection of stragglers. We suggest a machine-learning approach to address these problems, and introduce a ...
[Full version]

The majority of current methods in object classification use the one-against-res​t training scheme. We argue that when applied to a large number of classes, this strategy is problematic: as the number of classes increases, the negative class becomes a very large and complicated collection of images. The resulting classification problem then becomes extremely unbalanced, and kernel SVM classifiers trained on such sets require long training time and are slow in prediction. To address these problems, ...
[Full version]

For decades, our culture has been fascinated with the concept of the “Star Trek computer”. An all-knowing entity which is available to query about almost anything relevant to the world around us, organize our daily lives, remind us when we are doing the wrong thing, help us when we are lost. Several key features of this type of system are: (1) We can interact with it via natural language and it will understand our words ...
[Full version]

Due to the evolution of technology constraints, especially energy constraints which may lead to heterogeneous multicores, and the increasing number of transient faults and permanent defects, the design of defect-tolerant accelerators for heterogeneous multi-cores may become a major micro-architecture research issue. Most custom circuits are highly defect sensitive, a single transistor can wreck such circuits. On the contrary, neural networks (NNs) are inherently error-tolerant algorithms. And the emergence of high-performance applications implementing recognition, mining and ...
[Full version]

The CHILDES database is a large collection of child---adult spoken interactions in over 25 languages. Automatic annotation of these data faciliates research on child language development and acquisition by providing researchers with a large amount of accurate data. Recently, the English section of the CHILDES database was automatically annotated with labeled dependency relations in a state-of-the-art approach. We describe a similar endeavor, focusing on the Hebrew section of CHILDES. This is done by the following ...
[Full version]

We generalize sealed bid auctions to accommodate combinatorial auctions. In a sealed bid combinatorial auction each bidder sends to the auctioneer, simultaneously with the others, a message that depends only on his own valuation. The auctioneer decides on the allocation based on these messages alone. The goal is to find an allocation of the items which maximizes the social welfare. In this simultaneous communication complexity model we ask: How much information each of the bidders ...
[Full version]

Conformal maps are widely used in geometry processing applications. They are smooth, preserve angles, and are locally injective by construction. However, conformal maps do not allow for boundary positions to be prescribed. A natural extension to the space of conformal maps is the richer space of quasiconformal maps of bounded conformal distortion. Extremal quasiconformal maps, that is, maps minimizing the maximal conformal distortion, have a number of appealing properties making them a suitable candidate for ...
[Full version]

Dynamic events such as family gatherings, concerts or sports events are often captured by a group of people. The set of still images obtained this way is rich in dynamic content but lacks accurate temporal information. We propose a method for *photo-sequenci​ng* -- temporally ordering a set of still images taken asynchronously by a set of uncalibrated cameras. Photo-sequencin​g is an essential tool in analyzing (or visualizing) a dynamic scene captured by still images. We ...
[Full version]

On embedded systems, the Linux kernel doesn't have the BIOS to tell it what the hardware is like. On the other hand, the traditional solution of having the hardware information hardcoded in the kernel source is leading to an overpopulation of platform-specific hacks (read: a disaster).
[Full version]

Code breakers played an enormously crucial role in World War II. Alan Turing, the father of computer science, was at the center of allied code breaking operations and his breakthroughs made intelligence gathering not only possible but practical. This general audience talk, celebrating Alan Turing's Centenary, explains the notorious German Enigma code and how it was systematically cracked by the Allies. Bio Ymir Vigfusson is an Assistant Professor at the School of Computer Science at ...
[Full version]

NURBS is one of the typical way of representing geometric data and have been widely used in many areas such as computer graphics, CAGD, robotics. NURBS represents geometric data based on the mathematical form, and thus it requires relatively small memory space compared to the other representations. Compactness of geometric data is crucial in recent computing environments (Network, Mobile, Multi-Core) and NURBS is good candidate to be a geometric representation in such environments. However, at ...
[Full version]

The inverse RNA folding problem for designing sequences that fold into a given RNA secondary structure was introduced in the early 1990's in Vienna. Using a coarse-grain tree graph representation of the RNA secondary structure, we extended the inverse RNA folding problem to include constraints such as thermodynamic stability and mutational robustness, deveoping a program called RNAexinv. Furthermore, we propose a fragment-based design approach of RNA sequences that can be useful to practitioners in a ...
[Full version]

Given a function f: {0,1}^n \to {0,1}, the f-isomorphism testing problem requires a randomized algorithm to distinguish functions that are identical to f up to relabeling of the input variables from functions that are far from being so. An important open question in property testing is to determine for which functions f we can test f-isomorphism with a constant number of queries. Despite much recent attention to this question, essentially only two classes of functions ...
[Full version]

Distributed storage systems have become a popular solution to large file storage and fast data access. In such systems, erasure-correcting codes are widely used to combat disk failures, where disks correspond to symbols in the code. Specifically, we study MDS (maximum distance separable) array codes which enable optimal storage and efficient encoding and decoding algorithms. With r redundancy symbols an MDS code can sustain r erasures. For example, consider an MDS code that can correct ...
[Full version]

Effective segmentation prior to recognition has been shown to improve recognition performance. However, most segmentation algorithms adopt methods which are not explicitly linked to the goal of object recognition. Here we solve a related but slightly different problem in order to assist object recognition more directly - the extraction of a foreground ask, which identifies the locations of objects in the image. We propose a novel foreground/background segmentation algorithm that attempts to segment the interesting ...
[Full version]

The traditional performance-cost benefits enjoyed for decades from scaling of device area are challenged by the slowdown of voltage scaling and less reliable silicon primitives. These developments lead to pessimistic projections that it will be impossible to operate all on-chip resources, even at the minimum voltage for safe operation, due to power constraints, and the growing design and operational margins, used to provide silicon primitives with resiliency against variations, will consume the scaling benefits. Our ...
[Full version]

A rich man has many children. Unfortunately, over the years they have grew apart and started to hate each other. The rich man seeks to give his inheritance to the largest group of his children that can cooperate. How can he do it? To answer this, we introduce and study the related notions of lossy chains and fractional secret sharing. Both of these concepts are motivated by goal of controlling the amount of work required ...
[Full version]

A crucial step in the evolution of broadband wireless (cellular) networks is reducing the size of the cells and increasing their number. This target is usually obtained using cell sectorization, where the omni-directional antenna at each base station (BS) is replaced by 3 or 6 directional antennas. With respect to this evolution, the contribution of our work is two-fold. First, we propose a new protocol stack for a BS that governs multiple directional antennas. In ...
[Full version]

The introduction of a non-black-box simulation technique by Barak (FOCS 2001) has been a major landmark in cryptography, breaking the previous barriers of black-box impossibility. Barak’s techniques were subsequently extended and have given rise to various powerful applications. We present the first non-black-box simulation technique that does not rely on Barak’s technique. Our technique is based on essentially different tools: it does not invoke universal arguments, nor does it rely on collision-resistant hashing. Instead, the ...
[Full version]

We present the CBTree, a new counting-based self-adjusting sequential search tree that, like splay trees, moves more frequently accessed nodes closer to the root: After M operations on N items, Q of which access some item V, an operation on V traverses a path of length O(log M/Q) while performing few if any rotations. In contrast to the traditional self-adjusting splay tree in which each accessed item is moved to the root through a sequence ...
[Full version]

Covering questions emerges in many disciplines and are closely related to the well known set-cover problem in computer science. Similarly, geometric covering is of great importance and yet has only been investigated in seemingly unrelated specific disciplines. Examples include the well known art-gallery problem, mold-design problems, inspection, security and surveillance problems. In this thesis, we present a single unified framework that can solve many of the above geometric covering queries. The suggested framework reduces a ...
[Full version]

We describe a bootstrapping algorithm able to learn from partially labeled data. We report the results of an empirical study for using this algorithm to improve performance of sentiment classification using up to 15 million unlabeled Amazon product reviews. Our experiments cover semi-supervised learning, domain adaptation and weakly supervised learning. In some cases our methods were able to reduce test error by more than half using such large amounts of data, and in all cases ...
[Full version]

In the famous order maintenance problem, one wishes to maintain a dynamic list L of size n under insertions, deletions, and order queries. In an order query, one is given two nodes from L and must determine which node precedes the other in L. In an extension to this problem, named the Predecessor search on Dynamic Subsets of an Ordered Dynamic List problem (POLP for short), it is also necessary to maintain dynamic subsets S_1... ...
[Full version]

Finding dense intrinsic correspondence between non-rigid shapes is a notoriously difficult problem with many important applications in computer graphics and pattern recognition. In the first part of the talk, I will present a novel sparse modeling approach to non-rigid shape matching using only the ability to detect repeatable regions. As the input to our algorithm, we are given only two sets of regions in two shapes; no descriptors are provided so the correspondence between the ...
[Full version]

Texture features have always been a key attribute in image recognition and classification. In this work we propose a pre-processing stage for enhancing the performance of widely used color texture recognition methods. One approach we investigated, Decorrelation Stretching, was employed historically for enhancing the interpretability of multi-channel satellite images. This is achieved by stretching the dynamic range of color data over its principal components. Another approach decomposes the image into cartoon-like and texture components and ...
[Full version]

A numerical framework for the simulation of electrokinetic migration of particles in an electrolyte solution due to the application of an external electric field is presented. The electrokinetic transport process is described by a system of nonlinear partial differential equations (PDE). A thin boundary layer forms around the particle due to strong electrostatic forces. The resulting scale disparity of the boundary layer is used to derive nonlinear effective boundary conditions for the numerical solver using ...
[Full version]

In a distributed message passing model a communication network is represented by an n-vertex graph whose vertices host processors, and edges serve as communication links. One of the most fundamental goals in this setting is breaking the symmetry in the network. Specifically, the tasks of computing vertex coloring, maximal matching, and maximal independent set are of great importance. In the mid-eighties several randomized distributed algorithms for these problems were devised. [Luby 86, Alon, Babai and ...
[Full version]

Wireless networks are the most important assets in our industry, yet the problem is that we are building new networks the same way we built our previous networks. Capacity is planned in advance per service; to handle peek traffic and to add new capacity takes time and investment of dedicated equipment. We have no elasticity as demand grows, on average the service is grossly underutilized though also over utilized at specific times of day or ...
[Full version]

Recent studies indicate that multiple patches to software are found in a hefty portion of resolved bugs. It is also known that bugs that require multiple patches take longer to resolve, that their severity tends to be higher than the average and that they induce programmers to engage more in bug discussions. This work is concerned with the ability of programmers to predict a bug will be of this sort, and in particular that it ...
[Full version]

Optimal use of computing resources is pivotal to minimize expenditures and improve competitiveness of companies. The computing technologies are continuously evolving and the users are adapting at a unparalleled pace, thus producing intriguing research questions. In the first half of the talk, we observe that Internet service providers (ISPs) face surging Internet loads associated with real-time streaming video and various forms of dynamically generated, short-lived content. We propose a new system called GRADIENT aimed at ...
[Full version]

In the talk we present mathematical models and numerical methods which lead to early embryogenesis reconstruction and extraction of the cell lineage tree from the large-scale 4D image sequences. Robust and efficient finite volume schemes for solving nonlinear PDEs related to filtering, object detection and segmentation of 3D images were designed to that goal and studied mathematically. They were parallelized for massively parallel computer clusters and applied to the mentioned problems in developmental biology. The ...
[Full version]

Minimizing the discrepancy of a set system is a fundamental problem in combinatorics. One of the cornerstones in this area is the celebrated six standard deviations result of Spencer (AMS 1985): In any system of $n$ sets in a universe of size $n$, there always exists a coloring which achieves discrepancy $6\sqrt{n}$. The original proof of Spencer was existential in nature, and did not give an efficient algorithm to find such a coloring. Recently, a ...
[Full version]

As each technology generation brings additional transistors, the computer industry hopes to convert these into performance growth by stamping out a greater number of cores on a die. On the one hand, in many environments, that seems like a lot of hope. On the other hand, architecture researchers have grown almost allergic to "complex" alternatives, which history has shown can quickly fall off the cliff of diminishing returns. A fundamental hurdle to bettering architectures may ...
[Full version]

As each technology generation brings additional transistors, the computer industry hopes to convert these into performance growth by stamping out a greater number of cores on a die. On the one hand, in many environments, that seems like a lot of hope. On the other hand, architecture researchers have grown almost allergic to "complex" alternatives, which history has shown can quickly fall off the cliff of diminishing returns. A fundamental hurdle to bettering architectures may ...
[Full version]

The Digital Rights Movement stands between advanced technology and people rights (in the broad sense). Technological advance provides us with new products and new means of interacting with our surrounding. It provides a customer more flexibility and new ways of using a given product. However, it can be used to control and limit the usability of the product. For example, a customer buys a product, a certain book. Does it matters if the product is ...
[Full version]

We present several variants of the sunflower conjecture of Erdos and Rado and discuss the relations among them. We then show that two of these conjectures (if true) imply negative answers to questions of Coppersmith and Winograd and of Cohn et al regarding possible approaches for obtaining fast matrix multiplication algorithms. Specifically, we show that the Erdos-Rado sunflower conjecture (if true) implies a negative answer to the ``no three disjoint equivoluminous subsets'' question of Coppersmith ...
[Full version]

Automatic parallelization is a promising strategy to improve application performance in the multicore era. However, common programming practices such as the reuse of data structures introduce artificial constraints that obstruct automatic parallelization. Privatization relieves these constraints by replicating data structures, thus enabling scalable parallelization. Prior privatization schemes are limited to arrays and scalar variables because they are sensitive to the layout of dynamic data structures. This work presents Privateer, the first fully automatic privatization system ...
[Full version]

We develop a framework for proving lower bounds on computational problems over distributions, including optimization and unsupervised learning. Our framework is based on defining a restricted class of algorithms, called statistical algorithms, that instead of accessing samples from the input distribution can only obtain an estimate of the expectation of any given function on a sample drawn randomly from the input distribution. Our definition captures many natural algorithms used in theory and practice, e.g. moments-based ...
[Full version]

Kademlia is considered to be one of the most effective key based routing protocols. It is nowadays implemented in many file sharing peer-to-peer networks such as BitTorrent, KAD, and Gnutella. Kaleidoscope , a novel routing/caching scheme designed to significantly reduce the cost of Kademlia lookup operations Kaleidoscope also improves the anonymity of nodes in the system.
[Full version]

Mark Twain famously said that ``the past does not repeat itself, but it rhymes.'' In the spirit of this reflection, we present novel algorithms and methods for leveraging large-scale digital histories and human knowledge mined from the Web to make real-time predictions about the likelihoods of future human and natural events of interest. The Web is a dynamic being, with constantly updating content, which is entangled with sophisticated user behaviors and interactions. Some of these ...
[Full version]

Mark Twain famously said that ``the past does not repeat itself, but it rhymes.'' In the spirit of this reflection, we present novel algorithms and methods for leveraging large-scale digital histories and human knowledge mined from the Web to make real-time predictions about the likelihoods of future human and natural events of interest. The Web is a dynamic being, with constantly updating content, which is entangled with sophisticated user behaviors and interactions. Some of these ...
[Full version]

Mark Twain famously said that ``the past does not repeat itself, but it rhymes.'' In the spirit of this reflection, we present novel algorithms and methods for leveraging large-scale digital histories and human knowledge mined from the Web to make real-time predictions about the likelihoods of future human and natural events of interest. The Web is a dynamic being, with constantly updating content, which is entangled with sophisticated user behaviors and interactions. Some of these ...
[Full version]

Mark Twain famously said that ``the past does not repeat itself, but it rhymes.'' In the spirit of this reflection, we present novel algorithms and methods for leveraging large-scale digital histories and human knowledge mined from the Web to make real-time predictions about the likelihoods of future human and natural events of interest. The Web is a dynamic being, with constantly updating content, which is entangled with sophisticated user behaviors and interactions. Some of these ...
[Full version]

Mark Twain famously said that ``the past does not repeat itself, but it rhymes.'' In the spirit of this reflection, we present novel algorithms and methods for leveraging large-scale digital histories and human knowledge mined from the Web to make real-time predictions about the likelihoods of future human and natural events of interest. The Web is a dynamic being, with constantly updating content, which is entangled with sophisticated user behaviors and interactions. Some of these ...
[Full version]

Method overloading is a controversial language feature, especially in the context of Object Oriented languages, where its interaction with overriding may lead to confusing semantics. One of the main arguments against overloading is that it can be abused by assigning the same identity to conceptually different methods. This talk describes a study of the actual use of overloading in Java. To this end, we developed a taxonomy of classification of the use of overloading, and ...
[Full version]

All images have noise, but this noise may have undergone many distortions. Can we take any image, say, an olde photograph, and denoise it? This requires a good denoising method and an accurate noise estimator, both working for "any" image and "any" noise. I'll discuss both aspects, and particularly how to estimate a signal-dependen​t and scale-dependent noise. A prototype of the noise clinic is currently on line at http://dev.ipol​.im/~colom/ipol​_demo/noise_cli​nic/ (username : demo, password: demo). This ...
[Full version]

The field of Parameterized Complexity strives to solve intractable problems efficiently, via multivariate analysis of running time, as a function of both the input size n and a parameter k. Such analysis enables to show that some of these problems are \emph{fixed parameter tractable} (FPT); in other words, they can be solved in time f(k)* n^O(1). The rationale behind this approach is the observation that many real-life inputs have small parameter values. In this work ...
[Full version]

Emergence can be defined as the way complex systems and patterns arise out of a multiplicity of relatively simple interactions. Emergent behavior based systems are believed to be scalable and robust due to their non-reliance on a critical central element or fragile hierarchy. In the talk, two novel distributed algorithms based on emergent behavior will be presented and their scalability and robustness will be discussed.
[Full version]

Over the next few years, a new model of buying and selling cloud computing resources will evolve. Instead of providers exclusively selling server-equivalent virtual machines for relatively long periods of time (as done in today's IaaS clouds), they will increasingly sell individual resources (such as CPU, memory, and I/O resources) for a few seconds at a time. We term this nascent economic model of cloud computing the Resource-as-a-Service (RaaS) cloud, and we argue that its ...
[Full version]

In complex search tasks that utilize information from several data sources, it is often required to pose several basic search queries, join the answers to these queries, where each answer is given as a ranked list of items, and return a ranked list of combinations. Example: A tourist Alice wants to find a festive event that is related to classical music, a good Italian restaurant and an underground station that contains an elevator, such that ...
[Full version]

This lecture will discuss the challenges presented by Hebrew vowel marks (nikud) with respect to glyph-placement and kerning in (La)TeX and under Linux, as well as the various approaches to date to meet these challenges, and likewise the cultural aspects of their causes and implications.
[Full version]

In the last two decades patenting software relations invention in the US was a flourishing business. Recent ruling by Federal Court and its affirmation in the Supreme Court of the US may seriously affect this. We will review the current legal status of software related patents in the US in view of this ruling as well as relate to the legal status in Europe and Israel. The recently enacted America Invents Act (AIA) will also ...
[Full version]

Subdivision schemes are attractive methods for generating a smooth object from discrete data by repeating refinements. These schemes have many desirable properties such as fast convergence and smoothness of the generated objects. Therefore, subdivision schemes have gained popularity in recent years as an important tool in approximation theory, computer graphics, geometric design and computer aided design. I will start with a survey on basic subdivision schemes. Then, I will review fundamental results in the field, ...
[Full version]

At the CVLAB, EPFL, We have developed a multi-people tracking system. Our system is able to reliably track multiple people in a multi-camera setting. The obtained trajectories can be used for understanding individuals and group behavior. For example, we are currently involved in a project whose goal is to understand the behavior of basketball teams and players from multiple video cameras. Our system is composed of three core components: detection, identification and tracking. The detection ...
[Full version]

Nearest-neighbor queries, which ask for returning the nearest neighbor of a query point in a set of points, are important and widely studied in many fields because of a wide range of applications. In many of these applications, such as sensor databases, location based services, face recognition, and mobile data, the location of data is imprecise. We therefore study nearest neighbor queries in a probabilistic framework in which the location of each input point and/or ...
[Full version]

Git is well-known as the tool for collaborating in free software projects, and the Linux kernel in particular. What seems to be less known, is how useful it is even in a single developer scenario. This is a slideless, demostration-based talk, walking through the practical work with git as a most valuable tool for keeping a solo project under control, and helping in real-life situations. Git-savvy participants are also welcomed to share their experience.
[Full version]

Blind Source Separation (BSS) is a very applicable and well-studied problem. Most studies of the BSS problem assume the system to be time/position invariant, an assumption which assists the mathematical study, but is not guaranteed for real world situations. We present a method applicable to the case of underdetermined time/position varying mixing systems, where number of mixture observation is limited to be less than the number of sources and the system is changing with time/position ...
[Full version]

In the problem of secure multiparty computation (MPC) we have n parties who want to jointly evaluate a function f on their local inputs. An MPC protocol should allow the parties to correctly compute f while hiding the inputs from each other to the extent possible. An important complexity measure of MPC protocols is their round complexity. This talk will cover two questions related to the goal of minimizing the round complexity of MPC: * ...
[Full version]

We present a novel algorithm based on a cascade of regularization terms designed to induce implicit hierarchical sharing of information among related learning tasks. Our approach can be viewed as training and combining a set of diverse classifiers. Such a combination is known to improve accuracy. The diversity is achieved by inducing different levels of sharing among tasks. Our approach is designed for multi-task and multi-class learning scenarios. Enabling different levels of shared information is ...
[Full version]

Search engines available today render almost useless when you need to process Hebrew texts. Even effective open-source solutions like Lucene/Solr give up in despair when handed a Hebrew corpus to index. HebMorph is an open-source project with the ultimate goal of solving this problem, and in the best way possible. In-depth understanding of how search engines work triggers ideas, ideas become software parts, and those in turn are experimented with various state of the art ...
[Full version]

Gene translation is a central process in all living organisms. Thus, developing a better understanding of this complex process and its computational modeling have ramifications to every biomedical discipline. We develop computational models of gene translation and employ evolutionary systems biology approaches to study how its efficiency is encoded in the transcripts. The talk will include details about our models and several very recent discoveries about the way evolution shapes the translation efficiency of transcripts.
[Full version]

Randomized node selection is widely used in large-scale, distributed storage systems to both load balance chunks of data across the cluster and select replica nodes to provide data durability. We argue that while randomized node selection is great for load balancing, it fails to protect data under a common failure scenario. We present MinCopysets, a simple, general-purpose and scalable replication technique to improve data durability while retaining the benefits of randomized load balancing. Our contribution ...
[Full version]

Multiview structure recovery from a collection of images requires the recovery of the positions and orientations of the cameras relative to a global coordinate system. We present an approach to this problem that given feature correspondences across pairs of images uses the pairwise Essential Matrices to recover camera orientation by applying robust optimization using either spectral or semidefinite programming relaxations. Once the orientations are recovered we return to the point correspondences and use them directly ...
[Full version]

Today's scanning technologies, allow fast 3D scanning of urban scenes. Such rapid acquisition incurs imperfections: large regions remain missing, significant variation in sampling density is common, and the data is often corrupted with noise and outliers. Buildings often exhibit large scale repetitions and self-similarities. To consolidate the imperfect data, our key observation is that the same geometry, when scanned multiple times over reoccurrences of instances, allow application of a simple and effective non-local filters. We ...
[Full version]

This talk considers the problem of recognizing activities in surveillance video. Activities are high-level non-atomic semantic concepts which may have complex temporal structure. Activities are not easily identifiable using image features, but rather by the recognition of their composing events. Unfortunately, these composing events may only be observed up to a particular certainty. Many approaches to classification/recognition in computer vision rely on the availability of a large corpus of training data. This data can be ...
[Full version]

While it has been long recognized that genes are not randomly disposed along the genome, the degree to which the three dimensional (3D) structure of the genome influences the arrangement of genes has remained elusive. In particular, 'transcriptional-factories' are thought to lead to the positioning co-regulated genes in physical proximity, however there have also been reports to the contrary. Here we present evidence that the co-localization of yeast genes regulated by the same factor is ...
[Full version]

CORFU* is a novel storage cluster design that pools a farm of flash units and exposes it to clients as a single, global shared-log. CORFU utilizes flash to break the seeming tradeoff between consistency and performance, providing both strong consistency guarantees and high throughput. Our implementation is carried mostly by a client-side library, thus relieving the service from any IO bottlenecks; a CORFU cluster of 40 flash units can drive roughly 1 million 4-KByte ...
[Full version]

We study the bandwidth allocation problem (BAP) and the storage allocation problem (SAP) in bounded degree trees. In BAP, we are given a tree and a set of tasks. Each task consists of a path in the tree, a bandwidth demand, and a weight. Our goal is to find a maximum weight subset S of tasks such that, for every edge e, the total bandwidth of demand in S whose path contains e does not ...
[Full version]

The popularity of web-based mapping services such as Google Earth/Maps and Microsoft Virtual Earth (Bing), has led to an increasing awareness of the importance of location data and its incorporation into both web-based search applications and the databases that support them. In the past, attention to location data had been primarily limited to geographic information systems (GIS), where locations correspond to spatial objects and are usually specified geometrically. However, in the web-based applications, the location ...
[Full version]

How cloud-friendly is traditional virtualization? As a matter of fact, all traditional virtualization technologies predate cloud era. Traditional virtualization has major advantage, they are backward compatible on binary level, making it easy to run in cloud any existing pre-cloud application. But what if backward compatability is not a requirement? What about new post-cloud applications, that are developed specifically for cloud? Is traditional virtualization still a good platform for them?
[Full version]

Locally Decodable Code (LDC) is a code that encodes a message in a way that one can decode any particular symbol of the message by reading only a constant number of locations, even if a constant fraction of the encoded message is adversarially corrupted. In this talk we will show connection between LDC and a representation theory. We show that if there exists an irreducible representation (\rho, V) of a group G and q elements ...
[Full version]

This talk will discuss some system issues related to sharing virtual memory between CPU and GPU, which brings new challenges: GPU page faults, flashing a TLB on GPU, etc. We will also describe ESFIR - an SVM prototype, and XTHREAS - a new SVM programming models, which extends pthread semantics to GPU.
[Full version]

With the rise of the internet and its applications, it has became more and more common for unfamiliar parties to interact with each other. In order to help and encourage such interactions trust management systems were introduced. Those systems try to alleviate mistrust by gathering and processing statistical information about previous events. This thesis describes TrustPack, a trust management framework that provides trust management as a service, i.e., TrustPack is separated from any specific application ...
[Full version]

It is sometimes required to run the same application program on different platforms, from smartphones to PC's. The application may have to employ different User Interfaces (UI's) on different platforms, e.g. a UI designed for a large PC screen may not fit into the small screen of a smartphone. A different version of the application may therefore be needed for each platform kind. Kantorowitz and Lyakas introduced the use of platform-independent semantic user interfaces. As ...
[Full version]

Our world is known for its abundance of symmetric structures - in the fields of the animal kingdom, in astronomy, mathematics and chemistry, to name a few. The existence of symmetry in 3D shapes is of great interest when such applications as efficient storage, comparison and lookup are considered. Traditionally, only symmetries which are a composition of rotations and reflections were considered. These symmetries, termed extrinsic, have limited use in non-rigid shapes, as they are ...
[Full version]

TBA
[Full version]

Supervised learning, one of the core models of machine learning, is concerned with inducing rules from a given sample of examples. Active learning is a variant of this model in which the learning algorithm can choose which examples to learn from. The performance of such active learners is measured by the "speed" of learning: The number of examples they consume in order to produce a competitive solution. In this talk, I will present a novel ...
[Full version]

The challenging task of Byzantine self-stabilizing pulse synchronization requires that, in the presence of a minority of nodes that are permanently maliciously faulty, the non-faulty nodes must start firing synchronized pulses in a regular fashion after a finite amount of time, regardless of the initial state of the system. We study this problem under full connectivity in a model where nodes have local clocks of unknown, but bounded drift, and messages are delayed for an ...
[Full version]

We address the problem of robot navigation using natural visual features in a planar environment. The algorithm gets as an input a target, which is represented by a captured image from the targets pose, and a source image, captured from the starting position and should make an accurate move towards the target which can be decomposed as a rotation and translation on the plane. This algorithm can be used as a step in a higher ...
[Full version]

Discrete energy minimization is a ubiquitous task in computer vision, yet it is NP-hard in most cases. In this work we propose a multiscale framework for coping with the NP-hardness of discrete optimization. Our approach utilizes algebraic multiscale principles to efficiently explore the discrete solution space, yielding improved results on challenging energies for which current methods provide unsatisfactory approximations. In contrast to popular multiscale methods in computer vision that builds an image pyramid, our framework ...
[Full version]

Let M be a low-rank matrix. For vectors x,y, define the bilinear form f(x,y)=x^t M y. We study the question of reconstructing M from evaluations to f. Much of previous work allowed randomized evaluations, or a stronger query model (or both). We show how to, in an optimal number of 4nr measurements, efficiently reconstruct M from deterministically chosen queries to f. This can be seen as a (noiseless) generalization of compressed sensing, and we make ...
[Full version]

In this talk we introduce generic convex spaces of bounded distortion piecewise linear mappings of triangular meshes. It is shown how common geometric processing objective functionals can be restricted to these new spaces, rather than to the entire space of piecewise linear mappings, to provide a bounded distortion version of popular algorithms.
[Full version]

A significant portion of the energy dissipated in modern integrated circuits is consumed by the overhead associated with timing guardbands that ensure reliable execution. Timing speculation, where the pipeline operates at an unsafe voltage with any rare errors detected and resolved by the architecture, has been demonstrated to significantly improve the energy-efficiency of scalar processor designs. Unfortunately, applying the same timing-speculative approach to wide-SIMD architectures, such as those used in highly-efficient GPUs, may not provide ...
[Full version]

Although many algorithms for computing Euclidean Voronoi diagrams of point sites have been published, most of them are sequential in nature and hence cast inherent difficulties on the possibility to compute the diagrams in parallel. We present a new algorithm which enables the (combinatorial) computation of each of the Voronoi cells independently of the other ones. The algorithm is significantly different from previous ones and some of the ideas related to it are in the ...
[Full version]

Many monitoring tasks over distributed data streams can be formulated as a continuous query using a function that is defined over the global average of data vectors derived from the streams. The query will typically produce an alert when the value of the function crosses a predefined threshold. A fundamental problem in efficient scalable implementation of such threshold queries is that the data streams are distributed, sometimes over a wide geographical region. Moving all the ...
[Full version]

Many text processing tasks are based on estimating semantic relatedness between texts. For example, in information retrieval, relevancy of documents can be determined based on the semantic distance from the query. Recently, many algorithms have been developed for evaluating semantic relatedness based on a conceptual representation of the input 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 ...
[Full version]

Given a set of $n$ points in $\ell_{1}$, how many dimensions are needed to represent all pairwise distances within a specific distortion ? This dimension-distortion tradeoff question is well understood for the $ell_{2}$ norm, where $O((\logn)/\epsilon^{2})$ dimensions suffice to achieve $1+\epsilon$ distortion. In sharp contrast, there is a significant gap between upper and lower bounds for dimension reduction in $\ell_{1}$. A recent result shows that distortion $1+\epsilon$ can be achieved with $O(n/\epsilon^{2})$ dimensions. On the ...
[Full version]

Fast and reliable bundle adjustment is essential in many structure from motion (SfM) related applications such as mobile vision, augmented reality and robotics. In this talk an incremental and computationally efficient method for bundle adjustment will be presented. The method incorporates two key ideas to substantially reduce the involved computational cost, compared to a conventional bundle adjustment. First, the cost function is formulated in terms of multi-view constraints instead of re-projection equations, thereby eliminating the ...
[Full version]

Smartphones are increasingly ubiquitous, and many users carry multiple phones to accommodate work, personal, and geographic mobility needs. We present Cells, a virtualization architecture for enabling multiple virtual smartphones to run simultaneously on the same physical cellphone in an isolated, secure manner. Cells introduces a usage model of having one foreground virtual phone and multiple background virtual phones. This model enables a new device namespace mechanism and novel device proxies that integrate with lightweight operating ...
[Full version]

We consider the following problem, which arises in many database and web-based applications: Given a set $P$ of $n$ points in a high-dimensional space $\reals^d$ and a distance $r$, we want to report all pairs of points of $P$ at Euclidean distance at most $r$. We present two randomized algorithms, one based on randomly shifted grids, and the other on randomly shifted and rotated grids. The expected performance of both algorithms is of the form ...
[Full version]

Cellular networks are becoming more and more crucial to our daily lives and operators are seeking new technologies for increasing their bandwidth. Examples for such technologies are cell sectorization, fractional frequency reuse, and coordinated multipoint Tx/Rx. To take advantage of these new technologies, the scheduler logic at the base station needs to determine not only when to transmit each packet but also what modulation and coding scheme to use, which frequency reuse area's resources, and ...
[Full version]

We consider the problem of testing if a given function f is close to a n-variate degree d polynomial over the finite field 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 affine subspace is ...
[Full version]

This talk concerns the estimation of human attributes from 3D human pose and motion. We consider both physical attributes (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. Based on an extensive corpus of motion capture data, with physical and perceptual ground truth, we analyze the inference of subtle biologically-inspired attributes ...
[Full version]

Let F be the field of q elements, where q=p^l for prime p. Informally speaking, a polynomial source is a distribution over F^n sampled by low degree multivariate polynomials. In this paper, we 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 individual degree d and total degree D. When p, D and ...
[Full version]

Dynamically changing (reconfiguring) the membership of a replicated distributed system while preserving data consistency and system availability is a challenging problem. In this talk I will discuss this problem in the context of Primary/Backup clusters and Apache Zookeeper. Zookeeper is an open source system which enables highly reliable distributed coordination. It is widely used in industry, for example in Yahoo!, Facebook,Twitter, VMWare, Box, Cloudera, Mapr, UBS, Goldman Sachs, Nicira, Netflix and many others. A common ...
[Full version]

This lecture aims to provide a brief introduction to the InfiniBand™ architecture and programming using RDMA verbs. InfiniBand is an open standard, used in HPC (supercomputing) and data center environments for high performance connectivity. 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 ...
[Full version]

Distance oracle is a data structure that provides fast answers to distance queries. Recently, the problem of designing distance oracles capable of answering restricted distance queries, that is, estimating distances on a subgraph avoiding some forbidden vertices, has attracted a lot of attention. In this talk, we will consider forbidden set 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 ...
[Full version]

We study self-optimizing networks and data structures. 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 networks is motivated by trends in today's Internet like large data centers and peer-to-peer networks. In this talk I will preset the general model of the problem and some initial theoretical results on data centers ...
[Full version]

We describe an approach to category-level detection and viewpoint estimation for rigid 3D objects from single 2D images. In contrast to many existing methods, we directly integrate 3D reasoning with an appearance-based voting architecture. Our method relies on a nonparametric representation of a joint distribution of shape and appearance of the object class. Our voting method employs a novel parametrization of joint detection and viewpoint hypothesis space, allowing efficient accumulation of evidence. We combine this ...
[Full version]

See Hebrew announcement.
[Full version]

What is an optimal formula computing a given multivariate 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 integer representing the product-depth of the hidden ...
[Full version]

SIFT-like local feature descriptors are ubiquitously employed in such computer vision applications as content-based retrieval, video analysis, copy detection, object recognition, photo-tourism, and 3D reconstruction from multiple views. Feature descriptors can be designed to be invariant to certain classes of photometric and geometric transformations, in particular, affine and intensity scale transformations. However, real transformations that an image can undergo can only be approximately modeled in this way, and thus most descriptors are only approximately invariant ...
[Full version]

Back in the early years of the dynamic web, Perl was the de-facto choice of language. It was so ubique that 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 couple of years a new bread of frameworks appeared in the Perl world. They are all based on the PSGI ...
[Full version]

Locally-computable pseudorandom generators (PRGs) map n random input bits into m>n pseudorandom bits such that each of the m outputs depend on a small number of d inputs. While it is known that such generators are likely to exist for the case of small sub-linear stretch m=n+n^{0.9}, it is less clear whether achieving larger stretch is possible. The existence of such PRGs, which was posed as an open question in previous works, has recently gained ...
[Full version]

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 information, 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 ...
[Full version]

In this talk, we present techniques to enable green communications in future generation of wireless systems that will rely 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 efficiency. However, many of these techniques significantly increase system complexity and energy consumption. Escalating energy costs and environmental concerns have already ...
[Full version]

The main problem with contemporary Computed Tomography (CT) imaging is the high radiation dose absorbed by patients during the screening. Reducing this dose may result in poor quality imaging when using the popular fast and direct reconstruction techniques. On the other hand, iterative methods powered by statistical models of the scan are better-performing in such cases, but are also very slow. To bridge the gap between these two solutions, various signal processing techniques that augment ...
[Full version]

Grid computing environments have become mission-critical components in research and industry, offering sophisticated solutions to exploit large computing and storage resources across multiple geographic locations and administrative domains. Usually, such grid resources are non-dedicated or opportunistic, as a consequence users will utilize the resources following a "best effort" approach. However, many real-world supercomputing applications, such as computational fluid dynamics, weather forecasting, and complex system simulations, rely on coallocation of large numbers of reliable resources as ...
[Full version]

Most algebraic multigrid (AMG) methods define the coarse operators by applying the (Petrov-) Galerkin coarse approximation (GCA) where the sparsity pattern and operator complexity of the multigrid 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 aggressiveness of the coarsening, which affect the complexity of the hierarchy of operators and the overall rate of convergence. ...
[Full version]

The study of combinatorial problems with a submodular objective function has attracted much attention in recent years, and is partly motivated by the importance of such problems to combinatorial optimization, economics, and algorithmic game theory. A partial list of well-known problems captured by submodular maximization includes: Max-Cut, Max-DiCut, Max-k-Cover, Generalized-Assignment, several variants of Max-SAT and some welfare and scheduling problems. While classical works on submodular maximization problems are mostly combinatorial in nature, many recent results ...
[Full version]

The study of combinatorial problems with a submodular objective function has attracted much attention in recent years, and is partly motivated by the importance of such problems to combinatorial optimization, economics, and algorithmic game theory. A partial list of well-known problems captured by submodular maximization includes: Max-Cut, Max-DiCut, Max-k-Cover, Generalized-Assignment, several variants of Max-SAT and some welfare and scheduling problems. While classical works on submodular maximization problems are mostly combinatorial in nature, many recent results ...
[Full version]

Access to high-quality data drives our ability to ask questions and find answers. Our lab has a number of projects based on this premise. Proteomics data from experimental research in diabetes has enabled the discovery of two promising beta cell regulators, Netrin and Sem3a. We have built MOPED, a human and model organisms’ protein expression database, to provide a comprehensive summary of publicly available proteomics data. With MOPED, users can browse, sort, query, visualize, and ...
[Full version]

A new robust estimation algorithm, the generalized projection based M-estimator (gpbM) is proposed. The algorithm is general and can handle heteroscedastic data where every point in the estimation have a different covariance. Does not require the user to specify any (scale) parameters, and can be applied to multiple linear constraints for single and multi-carrier problems. The gpbM has three distinct stages: scale estimation, robust model estimation and inlier/outlier dichotomy. The model estimation stage can be ...
[Full version]

With the emergence of multicore architectures there is a growing need for automatic parallelization, that distributes sequential code into multi threaded code. OpenMP defines language extensions to C, C++, and Fortran for implementing multi-threaded shared memory applications. Generation of such extensions by the compiler relieves programmers from the manual parallelization process. OpenMP specification has been implemented in GCC, and is part of the standard release since version 4.2. In this talk we review the OpenMP ...
[Full version]

In cooperative storage caching, clients may access blocks directly from each other's caches. Previous studies treated all the cooperating caches as a single pool, maximizing overall system performance at the price of possibly degraded performance for individual clients. In light of the popularity of many P2P mechanisms, we re-evaluate the concept of cooperative caching, considering selfish clients that cooperate only if they benefit from the interaction. This is the first study that considers selfish clients ...
[Full version]

This paper introduces the Furthest Hyperplane Problem (FHP), which is an unsupervised counterpart of Support Vector Machines. Given a set of n points in R^d, the objective is to produce the hyperplane (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 results regarding FHP. We provide both lower and upper ...
[Full version]

Machine virtualization, where many virtual machines run on a single physical machine, has become an increasingly popular area of research and development in the last decade. Despite the introduction of hardware support for machine virtualization in commodity servers, many workloads still suffer from degraded performance when run in virtual machines. This degradation can be particularly acute when an unmodified virtual machine -- a VM that is not aware it is running in a virtual environment ...
[Full version]

In 1996, twelve years after making Aliyah to Israel, Danny Lewin graduated from the Technion and came to MIT to study algorithms. Over the next few years, he wrote a prize-winning Master’s Thesis on Consistent Hashing and co-founded Akamai Technologies, which today accelerates the delivery of over 250,000 web sites, including all of the top media and commerce brands on the Web. In this talk, we will describe some of Danny’s early work on consistent ...
[Full version]

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!
[Full version]

Online learning and competitive analysis are two widely studied frameworks for online decision-making settings. Despite the frequent similarity of the problems they study, there are significant differences in their assumptions, goals and techniques, hindering a unified analysis and richer interplay between the two. In this work, we provide 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 ...
[Full version]

Semidefinite programming is a fundamental problem in convex programming with numerous applications. In the field of combinatorial 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 ...
[Full version]

Verb subcategorization frames determine the number and types of the syntactic arguments that verbs select, or subcategorize for. Typically, verbs can be associated with subcategorization frames that specify, for each subcategorized argument, information on the phrases that can realize this argument. These can be noun phrases (in the case of direct objects), prepositional phrases with one or more specific preposition, infinitival verb phrases, clauses introduced by a complementizer, etc. We introduce a technique and an ...
[Full version]

Sampling energy landscape of macromolecular interactions is a challenging problem of computational biology. Classical approaches like Molecular Dynamics and Monte Carlo are not always optimal for these applications due to large configurational space of the problem. On contrary The Fast Fourier Transform (FFT) correlation sampling approach can exhaustively evaluate the energies of billions of macromolecular configuratins on a grid provided two limitations - the energy is described in the form of a correlation function and ...
[Full version]

Given two monotone polynomials f,g, their unifier 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 monotone polynomials f,g which have a unifier, they can be computed by a small monotone 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 ...
[Full version]

Our visual attention is attracted by salient stimuli in our environment and affected by primitive features such as orientation, color, and motion. Perceptual saliency due to orientation contrast has been extensively demonstrated in behavioral experiments with humans and other primates and is commonly explained by the very particular functional organization of the primary visual cortex. We challenge this prevailing view by studying orientation-based visual saliency in two non-mammalian species with enormous evolutionary distance to humans. ...
[Full version]

Reconfigurable computing allows a lower power consumption 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 computational elements whose functionality is determined through multiple programmable configuration bits. The hardware can be re-programmed to implement an entirely different circuit. The focus of the research project ERA (Embedded Reconfigurable Architectures) is to investigate and propose new methodologies in both ...
[Full version]

PCI Express is an industry standard, and the most prominent interconnection architecture for I/O devices and other boards inside a computer. It is defined and updated by the PCI-SIG(R) industry standards body, originally formed in 1992 as the Peripheral Component Interconnect (PCI) special interest group (SIG). Agilent Technologies, a prominent developer and provider of test equipment at all levels yet not a competitor in the computer market, has played a key role in the advancement ...
[Full version]

Traditional models of bendable surfaces are based on the exact or approximate invariance to deformations that do not tear or stretch the shape, leaving intact an intrinsic geometry associated with it. Intrinsic geometries are typically defined by shortest paths also known as geodesic distances, or diffusion processes on the surface like diffusion distances. Both methods are implicitly derived from the metric induced by the ambient Euclidean space. Here, we depart from this restrictive assumption by ...
[Full version]

This thesis presents quantum analogues of artificial neural networks. We analyze and compare their performance to known classical and previously proposed quantum models. First we propose a model for associative memory based on a modification of Grover's quantum search algorithm and prove that the capacity of the model is exponential in the number of bits. We present algorithms for pattern completion and correction and prove that the model does not suffer from spurious memories and ...
[Full version]

Computing environments become increasingly parallel, and it seems likely that we will see more cores on tomorrow's desktops and server platforms. In a highly parallel system, tracing garbage collectors may not scale well due to deep heap structures that hinder parallel tracing. In this work we start by analyzing which data structures make current Java benchmarks create deep heap shapes. It turns out that the problem is manifested mostly with benchmarks that employ queues and ...
[Full version]

Threshold monitoring applications in distributed stream networks contin- uously monitor the global score of the network and alert whenever a given threshold is crossed. The network's global score is computed by applying a certain scoring function over the aggregated 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 ...
[Full version]

Coordinating the proper ordering of events across remote sites is a central task of distributed applications. In asynchronous systems, such coordination depends in an essential way upon message chains, as captured by Lamport's happened-before relation. The relation provides a useful approximation of causality, in the sense that in asynchronous systems two event can only be causally related if they are Lamport related. The talk will consider coordination and causality in synchronous systems, where upper bounds ...
[Full version]

After decades of research, vaccines against some of the greatest viral threats are still lacking and antiviral drugs remain few and slow in coming. These shortcomings point to the need for new approaches that go beyond traditional virology methods. High-throughput technologies and computational biology promise to deliver a much-needed boost to the field. My laboratory is using systems biology and computational approaches to understand and model integrated views of virus-host interactions, viral evasion of host ...
[Full version]

Providing higher performance than state-of-the-art embedded processors can deliver today will increase safety, comfort, number and quality of services, while also lowering emissions as well as fuel demands for automotive, avionic and automation applications. Engineers who design hard real-time embedded systems in such embedded domains express a need for several times the performance available today while keeping safety as major criterion. A breakthrough in performance is expected by parallelising hard real-time applications and running them ...
[Full version]

Emerging computer architectures pose many new challenges for software development. First, as the number of computing elements constantly increases, the importance of scalability of parallel programs becomes more significant. Second, accessing memory has become the principal bottleneck, while multi-CPU systems are based on NUMA architectures, where memory access from different chips is asymmetric. Therefore, it is important to design software with local data access, cache-friendliness, and reduced contention on shared memory locations, especially across chips. ...
[Full version]

In recent years, stochastic approximation approaches such as Stochastic Gradient Descent (SGD) and Stochastic Dual Averaging have become the optimization method of choice for many learning problems, including linear Support Vector Machines (SVMs). This is not surprising, since such methods yield optimal generalization guarantees with only 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 ...
[Full version]

Data centers are becoming the hosting platform for a wide spectrum of composite applications. In recent years, large investments have been made in massive data centers supporting cloud services, by companies such as eBay, Facebook, Google, Microsoft, and Yahoo!. With an increasing trend towards global and more communication intensive applications, the bandwidth usage within and between data centers is rapidly growing. The placement of the data used by these global applications presents a challenging optimization ...
[Full version]

Sparse representation theory has been increasingly used in signal processing and machine learning. In this work we introduce a hierarchical sparse modeling approach which integrates information from the image patch level to derive a mid-level invariant image and pattern representation. The proposed framework is based on a hierarchical architecture of dictionary learning for sparse coding in a cortical (log-polar) space, combined with a novel pooling operator which incorporates the Rapid transform and max pooling to ...
[Full version]

LaTeX is a high-quality typesetting system; BEAMER 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 comprehensive Hebrew presentations using Beamer in minutes. The talk assumes prior knowledge in Hebrew, but not in LaTeX nor Beamer. We will cover basic concepts, commands and examples in ...
[Full version]

We deal with the Regenerator Location Problem in optical networks. We are given a network G = (V, E), and a set Q of communication requests between pairs of terminals in V. We investigate two variations: one in which we are given a routing P of the requests in Q, and one in which we are required to find also the routing. In both cases, each path in P must contain a regenerator after every ...
[Full version]

A basic requirement in many distributed systems is the ability to detect objects whose score, according to a given function, exceeds some threshold. Since an object's data can be partitioned over various nodes, computing its global score requires collecting its data over the network. A main challenge is to perform threshold queries or monitoring with minimum network communication, i.e., without collecting the data from the nodes to a central location. In this talk I will ...
[Full version]

Efficient concurrent data structures are extremely important for obtaining good performance for most parallel programs. However, ensuring the correctness of concurrent data structure implementations 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 implementation of concurrent data structures is even more complex because data structure implementations must also detect conflicts between concurrent ...
[Full version]

Personalized treatment approaches which optimize drugs doses according to pre-treatment and early response-to-therapy evaluation hold the promise to improve treatment success rates and reduce severe adverse side-effects due to drugs toxicity in variety of pathologies. Reliable assessment of tissue microenvironment including cell proliferation, density and size and tissue perfusion as a biomarker for disease activity is a key necessity for personalized, response-based treatment regimes. Histology-based tissue microenvironment analysis requires invasive, surgical procedure to obtain the ...
[Full version]

C++ AMP (Accelerated Massive Parallelism) is a native programming model that contains elements that span the C++ programming language and its runtime library. The syntactic changes introduced by AMP are minimal, but additional restrictions are enforced to reflect the limitations of data parallel hardware. Data parallel algorithms are supported by the introduction of multi-dimensional array types, array operations on those types, indexing, asynchronous memory transfer, shared memory, synchronization and tiling/partitioning techniques. The language aims to ...
[Full version]

Contemporary mobile devices are equipped with multiple wireless interfaces, such as WiFi, Bluetooth, WiMax, ZigBee, NFC, etc. All these technologies differ dramatically one from another in maximum transmission range, bandwidth and power demands. Among all subsystems operating inside mobile devices, wireless communication is known as being particularly power-hungry, accounting for 50-70% of the total power consumption in small handheld devices, such as smartphones, and for 10% in laptops. Given the varying characteristics of different wireless ...
[Full version]

Much effort has been directed at algorithms for obtaining the highest probability (MAP) configuration in a probabilistic (random field) model. In many situations, one could benefit from additional solutions with high probability. Current methods for computing additional most probable configurations produce solutions that tend to be very similar to the MAP solution and each other. This is often an undesirable property. I will describe an algorithm for the M-Best Mode problem, which involves finding a ...
[Full version]

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 research 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 "isolating") the morphological properties of words to discover potential MWEs and show that this method improves the alignment ...
[Full version]

It is well known (c.f., Impagliazzo and Luby [FOCS '89]) that the existence of almost all ``interesting" cryptographic applications, i,e., ones that cannot hold information theoretically, implies one-way functions. An important exception where the above implication is not known, however, is the case of coin-flipping protocols. Such protocols allow honest parties to mutually flip an unbiased coin, while guaranteeing that even a cheating (efficient) party cannot bias the output of the protocol by much. While ...
[Full version]

Audio denoising is a long studied problem, with numerous algorithms and a wide accumulated knowledge. Considering non-stationary noise (like noise in a cocktail party environment) and strong, this task becomes very difficult to handle. This paper considers such audio denoising problems, where the audio track is accompanied by a video. Furthermore, the disturbing source is generally not visible in the field of view, and its nature is unknown. Overcoming such unknown noise is different than ...
[Full version]

The $k$-server problem is one of the most fundamental and extensively studied problems in online computation. Suppose there 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 given 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 ...
[Full version]

In the distributed average consensus problem, each node in a network has an initial value, and the objective is for all nodes to reach consensus at the average of these values using only communication with nearby nodes. Distributed average consensus algorithms have a wide variety of applications, including distributed optimization, sensor fusion, load balancing, and autonomous vehicle formation control. This talk centers on the analysis of distributed averaging algorithms for consensus and vehicle formation control ...
[Full version]

Why do we perceive some elements in a visual scene, while others remain undetected? We compared fixations on detected vs. undetected items in the Identity Search Task (Jacob & Hochstein, 2009). Using a gaze-contingent technique, we further controlled the number of fixations on the target (Jacob & Hochstein, 2010). Results show that detected targets were fixated at a greater extent, and a backward dynamics alignment revealed a bifurcation point where the differential characteristics begin. Moreover, ...
[Full version]

The complexity of several prominent graph polynomials, such as the chromatic polynomial and the matching polynomial, has been studied in the literature. In 2008 Makowsky raised a conjecture which generalizes complexity results for specific graph polynomials to an infinite class of graph polynomials which include almost all of those in the literature. The conjecture states roughly that the evaluations of such graph polynomials are equivalent in terms of running-time complexity, except for a small and ...
[Full version]

I will present a new approach for creating chosen ciphertext secure encryption. The focal point of our work is a new abstraction that we call Detectable Chosen Ciphertext Security (DCCA). Intuitively, this notion is meant to capture systems that are not necessarily chosen ciphertext attack (CCA) secure, but where we can detect whether a certain query CT can be useful for decrypting (or distinguishing) a challenge ciphertext CT*. We show how to build chosen ciphertext ...
[Full version]

Open Shortest Path First (OSPF) is the most popular interior gateway routing protocol on the Internet. Most known OSPF attacks that have been published in the past are based on falsifying the link state advertisement (LSA) of an attacker-controlled router. These attacks 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 ...
[Full version]

We presents a novel approach to pose estimation and model-based recognition of specular objects in difficult viewing conditions, such as low illumination, cluttered background, and large highlights and shadows that appear on the object of interest. In such challenging conditions conventional features are unreliable. We show that under the assumption of a dominant light source, specular highlights produced by a known object can be used to establish correspondence between its image and the 3D model, ...
[Full version]

In the last two years a committee in the Standards Institution of Israel worked, on my initiative and with my active participation, on a revision for the standard of the Hebrew keyboard layout. Why did i want to change the layout? Why is it good to do it through the Standards Institution? How to make people who represent Microsoft, Linux, and Apple agree about things? How is a keyboard layout edited in Linux and in ...
[Full version]

We describe an explicit, simple, construction of large subsets of F^n, where F is a finite field, that have small intersection with every k-dimensional affine subspace. Interest in the explicit construction of such sets, termed 'subspace-evasive' sets, started in the work of Pudlak and Rodl (2004) who showed how such constructions over the binary field can be used to construct explicit Ramsey graphs. More recently, Guruswami (2011) showed that, over large finite fields (of size ...
[Full version]

Nuclear magnetic resonance (NMR) has proven to be a leading implementation of quantum information processors where each molecule in the sample constitutes a register of quantum bits (qubits). However, 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 coding (compression) can cool some spins (reducing their Shannon entropy) while heating others, yet this closed-system technique is limited by Shannon's entropy ...
[Full version]

We show how to transform any additively homomorphic private-key encryption scheme that is compact, into a public-key encryption 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 distribution of homomorphically generated encryptions (in particular, we do not require them to be distributed like real ciphertexts). Our resulting public-key scheme is homomorphic ...
[Full version]

Graph matching is the problem of matching nodes of a graph in pairs such that the largest number of pairs is created or the largest sum of edge weights is obtained. Greedy graph matching provides us with a fast way to coarsen a given graph during graph partitioning. Direct 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 ...
[Full version]

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 messaging, presence information, file transfer and online games. This work introduces several novel load balancing algorithms for distributing SIP requests to a cluster of SIP servers. Our load balancer improves both throughput and response time versus a single node, while exposing a single interface to external clients. We present ...
[Full version]

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 images of the same scene taken under different lighting conditions, our method generates a joint segmentation of the images. We introduce a novel contour-based representation that allows us to cast the shape-based joint clustering problem as a quadratic semi-assignment problem. Our score function is additive. We use ...
[Full version]

This is a 4 series of 4 talks about GPGPUS, intended 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 became a hot topic 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 series of lectures ...
[Full version]