קולוקוויום וסמינרים

כדי להצטרף לרשימת תפוצה של קולוקוויום מדעי המחשב, אנא בקר בדף מנויים של הרשימה.

קולוקוויום וסמינרים בקרוב

• Hebrew Acronyms: Identification, Expansion, and Disambiguation

דובר:
קיילה ג'קובס, הרצאה סמינריונית למגיסטר
תאריך:
יום שני, 9.12.2013, 16:30
מקום:
טאוב 601
מנחה:
Prof. Alon Itai and Prof. Shuly Wintner

Acronyms are words formed from the initial letters of a phrase. For example, "CIA" is an acronym that usually means "Central Intelligence Agency," but in other contexts could mean "Culinary Institute of America." Understanding acronyms is important for many natural language processing applications, including search and machine translation. While hand-crafted acronym dictionaries exist, they are limited and require frequent updates. We developed a new machine learning method to automatically build a Hebrew acronym dictionary from unstructured text documents. This is the first acronym dictionary construction technique, in any language, to specifically include acronyms whose expansions do not necessarily appear in the same document. We also enhanced the dictionary with contextual information to help during acronym disambiguation. Additionally, while acronyms have a long history in Hebrew, and have previously been investigated from a qualitative linguistic perspective, they have never before been studied quantitatively. We'll share some new statistically-based linguistic insights about acronym usage in modern Hebrew texts, of interest to Hebrew language aficionados and developers of Hebrew natural language processing systems.

• Pixel Club: Medical Imaging and Computer Vision @IBM.Haifa

דובר:
פאבל קיסילב (י.ב.מ. חיפה)
תאריך:
יום שלישי, 10.12.2013, 11:30
מקום:
חדר 337, בניין טאוב למדעי המחשב

In this talk we present various research topics in medical imaging and related to them computer vision methods that are developed at IBM Haifa Research Lab. In the first part of the talk we describe various medical imaging problems and briefly hint on their solutions. The second part of the talk will be devoted to a newly developed general method for figure-ground segmentation. The method combines a bottom up approach of generating multiple highly plausible labelings of an image with a top-down model for reranking these segmentations, giving competitive results on a number of challenging datasets. Contrary to competing state-of-the-art methods for figure/ground segmentation that rely on complex models of foreground and boundary, our approach uses a simple binary pairwise graphical-model for bottom-up inference which relies on features computed on superpixels in the image and can be learned efficiently. The ranking stage, trained using features that rely on both the image evidence and inferred segmentation, discriminates between good and bad figure/ground labelings produced from the bottom-up inference. We will present examples of application of the proposed segmentation method to natural images and to difficult medical image cases.

• Theory Seminar: Playing Non-linear Games with Linear Oracles

דובר:
דן גרבר (טכניון)
תאריך:
יום רביעי, 11.12.2013, 12:30
מקום:
טאוב 201

Online Learning deals with playing repeated games against an adversary with the aim of minimising a quantity known as regret, and has attracted much attention in recent years due to its applications in algorithm design and machine learning. The computational bottleneck in the application of online learning algorithms is the computation of a projection of a point onto a convex set, which for many problems of interest, such as those that arise in combinatorial optimization and matrix optimization, is prohibitive. An alternative approach is to design algorithms for online learning that trade expensive projection steps in favour of linear optimization steps over the feasible domain, which for many problems are much more efficient and simple to implement.

In this talk we will present an algorithm for online learning that uses only linear optimization steps over the feasible domain (one per iteration of the game) and attains optimal regret, resolving an open question by Vempala and Kalai (COLT 2003), and Hazan and Kale (ICML 2012), in case the feasible domain is a polytope.

Our method is based on a novel variant of the conditional gradient method, that reduces the task of minimizing a smooth convex function over a domain to that of minimizing a linear objective. Whereas previous variants of this method give rise to approximation algorithms, we give such algorithm that converges exponentially faster.

The talk is based on a joint work with Elad Hazan (FOCS '13).

• Randomized Algorithms for Faster Linear and Non-linear Regression

דובר:
Haim Avron - CS-Lecture - Note unusual day
תאריך:
יום רביעי, 11.12.2013, 14:30
מקום:
חדר 337-8 טאוב.
קישור:
http://www.cs.technion.ac.il/~colloq/20131211_14_30_Avron.html
• Rank Modulation for Flash Memory

דובר:
מיכל הורוביץ, הרצאה סמינריונית למגיסטר
תאריך:
יום חמישי, 12.12.2013, 11:30
מקום:
טאוב 701
מנחה:
Prof. Tuvi. Etzion

Snake-in-the-box code is a Gray code which is capable to detect a single error. Gray codes are important in the context of rank modulation scheme which was suggested recently for representing information in flash memories. For a Gray code in this scheme the codewords are permutations, two consecutive codewords are obtained by using the "push-to-the-top" operation, and the distance measure is defined on permutations. In this talk the Kendall's $\tau$-metric is used. We present a general method for constructing such Gray codes. We apply the method recursively to obtain a snake of length $((2n+1)2n-1) M$ for permutations of $S_{2n+1}$ from a snake of length $M$ for permutations of $S_{2n-1}$. We present a direct construction based on necklaces which might yield snakes of length $\frac{(2n+1)!}{2}-2n+1$ for permutations of $S_{2n+1}$. The construction is applied successfully for $S_7$ and $S_9$.

• CGGC Seminar: Multi-view Inter-media: From Space to Ocean-depths

דובר:
יואב י. שכנר (הנדסת חשמל, טכניון)
תאריך:
יום ראשון, 15.12.2013, 13:00
מקום:
חדר 337, בניין טאוב למדעי המחשב

This talk is about ;multi-view imaging via participating media, particularly the 3D volumetric scattering atmosphere and the 3D wavy water-air interface. Camera multi-views are either looking down from outer-space, or looking up from underwater or the ground level.  Multiple views on scales of tens or hundreds of kilometers effectively create a huge lightfield camera-system. This provides constraints for recovering complex scenes, including the 3D distribution of aerosols overhead, or underwater topography. On a small scale, multiple submerged cameras viewing via random water-surface waves create a 'virtual periscope': it stochastically localizes objects in open air, without revealing the viewer. This may have implications to some marine animals.

See more at:

• כנס T2med בנושא: Social-Mobile-Cloud Meets Medicine @ Technion

תאריך:
יום שלישי, 17.12.2013, 08:30
מקום:
אודיטוריום קוגן, בניין מאייר, הפקולטה להנדסת חשמל

הנכם מוזמנים לכנס T2med בנושא: Social-Mobile-Cloud Meets Medicine @ Technion אשר יתקיים ביום ג', 17 בדצמבר 2013.

ההשתתפות מותנית בהרשמה מוקדמת. פרטים נוספים ותוכנית הכנס באתר הכנס.

• Pixel Club: Video Saliency and its Applications in Single and Multi-camera Setups

דובר:
דמיטרי רודוי (יפורסם)
תאריך:
יום שלישי, 17.12.2013, 11:30
מקום:
חדר 1061, בניין מאייר, הפקולטה להנדסת חשמל

Understanding human attention have interested researchers for decades. The early works come from different fields of psychology and separate the cognitive process into several steps. The different models of attention in static scenes have emerged and evolved into dynamic saliency. Along with that, there are extensive cinematographic theories on how the scene should be watched, or filmed. And again, there is a long term research interest in view selection for static and dynamic scenes. Different methods propose how to place a camera in a scene and how to move it. The central contribution of this research is a novel approach to video saliency modeling. We propose a model that can effectively predict humans' attention in any particular video. The system is learn from human examples, so our second contribution is an effective method for massive collection of gaze data. We adapt our model to multiple camera scenarios by proposing an approach for view selection based on fixed cameras. As the last contribution we propose a method to shift human attention by inlaying artificial objects into a video. Our model for video saliency is based on modeling gaze as attention shifts between consecutive video frames. This is different from analyzing each image independently, as was often done before and allows us to maintain temporal stability of the saliency maps. We incorporate static, motion and semantic features from the video to propagate a saliency map from one frame to another. We show that this model is better to the behavior of the human eyes. Since our saliency model is learn from a large database of human gaze tracks we additionally propose a method to collect them from any number of participants. The method employs crowdsourcing technique and allows to record gaze location on any number of frames of any video. Opposite to the traditional gaze tracking methods, our method does not require any special equipment and participants are not limited by any geography or culture.

As an approach to multiple camera setups we propose a method for efficient viewpoint selection from any set of cameras that view the same scene. As placing a camera at specified place usually requires knowledge of 3D data our method works with fixed cameras. It is capable of ranking the cameras according to the visibility of the actions happening in the scene. After the best view is selected the video saliency method can be applied to the resulting set of frames.

We further wish to edit the input video and shift the humans' attention. To do so we propose a user-friendly system for seamless inlaying of any 3D object into any video. We model the video as a single image, ask the user to add the object in the desired place and then render it back to the video.

To verify the proposed methods we test them on known video datasets and on real-life videos. We compare our results quantitatively to the state-of-the-art methods and outperforms them. Additionally, we present qualitative tests that render our results more visually appealing that the previous approaches.

• Predicate Abstraction For Relaxed Memeory Model

דובר:
יורי משמן, הרצאה סמינריונית למגיסטר
תאריך:
יום שלישי, 17.12.2013, 11:30
מקום:
טאוב 601
מנחה:
Prof. E. Yahav

In this talk we present a novel approach for predicate abstraction of programs running on relaxed memory models. Our approach consists of two steps. First, we reduce the problem of verifying a program P running on a memory model M to the problem of verifying a program P_M that captures an abstraction of M as part of the program. Second, we show how to discover new predicates that enable verification of P_M. The core idea is to extrapolate from the predicates used to verify P under sequential consistency. A key new concept is that of cube extrapolation: it successfully avoids exponential explosion when abstracting P_M. We implemented our approach for the x86 TSO and PSO memory models and showed that predicates discovered via extrapolation are powerful enough to verify several challenging concurrent programs. We show that our cube extrapolation approach introduces a x100 reduction in the number of SMT calls for several programs, enabling their verification. This is the first time some of these programs have been verified for a model as relaxed as PSO.

• New Approaches to Graph Partitioning

דובר:
Roy Schwartz - CS-Lecture
תאריך:
יום שלישי, 17.12.2013, 14:30
מקום:
חדר 337-8 טאוב.
קישור:
http://www.cs.technion.ac.il/~colloq/20131217_14_30_Schwartz.html
• Practical Concurrent Binary Search Trees via Logical Ordering

דובר:
דנה דרקסלר, הרצאה סמינריונית למגיסטר
תאריך:
יום רביעי, 18.12.2013, 12:30
מקום:
טאוב 701
מנחה:
Prof. E. Yahav

In this talk, we present novel practical concurrent binary search tree (BST) algorithms based on the following general idea: we explicitly maintain logical ordering information in the data structure, permitting clean separation from its physical tree layout. We capture logical ordering using intervals, enabled by the following key observation: any set of totally ordered items can be uniquely partitioned into logical interval ranges with the property that an item belongs to the set if and only if the item is an endpoint of some interval. These ideas enable us to construct efficient, synchronization-free and intuitive lookup operations. Specifically, we present: (i) A concurrent non-balanced BST with a lock-free lookup, and (ii) A concurrent AVL tree with a lock-free lookup which proceeds without synchronizing with any mutating operation, including balancing operations.

• A Relational Framework for Information Extraction

דובר:
Benny Kimelfeld - CS-Lecture - Note unusual day
תאריך:
יום רביעי, 18.12.2013, 14:30
מקום:
חדר 337 טאוב.
קישור:
http://www.cs.technion.ac.il/~colloq/20131218_14_30_Kimelfeld.html
• יום עיון לרגל פרישתו לגימלאות של פרופ' שמואל כץ

תאריך:
יום חמישי, 19.12.2013, 08:30
מקום:
חדר 337, בניין טאוב למדעי המחשב

לרגל פרישתו לגימלאות של פרופ' שמואל כץ, יתקיים בפקולטה למדעי המחשב בטכניון יום עיון בנושא: "Aspects of Formal Verification".

הכנס יתקיים ביום ה', 19 בדצמבר 2013, בחדר 337, בניין טאוב למדעי המחשב.

ההשתתפות אינה כרוכה בתשלום אולם מחייבת הרשמה מראש.

כולם מוזמנים.

פרטים נוספים ותוכנית.

• Polynomial Bounds for the Grid-Minor Theorem

דובר:
Julia Chuzhoy - CSpecial Lecture-Note unusual time and place
תאריך:
יום חמישי, 19.12.2013, 13:30
מקום:
חדר 601 טאוב.
קישור:
http://www.cs.technion.ac.il/~colloq/20131219_13_30_Chuzhoy.html
• Shortest Path Queries: Static, Dynamic and Fault-tolerant.

דובר:
Shiri Chechik - CS-Lecture
תאריך:
יום ראשון, 29.12.2013, 14:30
מקום:
חדר 337-8 טאוב.
קישור:
http://www.cs.technion.ac.il/~colloq/20131229_14_30_Chechik.html
• Toward Better Depth Lower Bounds: An Information Complexity Approach to the KRW Composition Conjecture

דובר:
Or Meir
תאריך:
יום שלישי, 31.12.2013, 14:30
מקום:
חדר 337-8 טאוב.
קישור:
http://www.cs.technion.ac.il/~colloq/20131231_14_30_Meir.html
• On kinetic Delaunay triangulations

דובר:
Natan Rubin - CS-Lecture
תאריך:
יום ראשון, 5.1.2014, 14:30
מקום:
חדר 337-8 טאוב.
קישור:
http://www.cs.technion.ac.il/~colloq/20140105_14_30_Rubin.html