Colloquia and Seminars

To join the email distribution list of the cs colloquia, please visit the list subscription page.


Computer Science events calendar in HTTP ICS format for of Google calendars, and for Outlook.

Academic Calendar at Technion site.

Upcoming Colloquia & Seminars

  • CGGC Seminar: Consistent Functional Cross Field Design for Mesh Quadrangulation

    Speaker:
    Omri Azencot (CS, Technion)
    Date:
    Sunday, 25.6.2017, 14:00
    Place:
    Room 337 Taub Bld.

    We propose a novel technique for computing consistent cross fields on a pair of triangle meshes given an input correspondence, which we use as guiding fields for approximately consistent quadrangulations. Unlike the majority of existing methods our approach does not assume that the meshes share the same connectivity or even have the same number of vertices, and furthermore does not place any restrictions on the topology (genus) of the shapes. Importantly, our method is robust with respect to small perturbations of the given correspondence, as it only relies on the transportation of real-valued functions and thus avoids the costly and error-prone estimation of the map differential. Key to this robustness is a novel formulation, which relies on the previously-proposed notion of power vectors, and we show how consistency can be enforced without pre-alignment of local basis frames, in which these power vectors are computed. We demonstrate that using the same formulation we can both compute a quadrangulation that would respect a given symmetry on the same shape or a map across a pair of shapes. We provide quantitative and qualitative comparison of our method with several baselines and show that it both provides more accurate results and allows to handle more general cases than existing techniques.

  • Pixel Club: InterpoNet, A Brain Inspired Neural Network for Optical Flow Dense Interpolation

    Speaker:
    Shay Zweig (Bar-Ilan & Tel-Aviv University)
    Date:
    Tuesday, 27.6.2017, 11:30
    Place:
    EE Meyer Building 1061

    Sparse-to-dense interpolation for optical flow is a fundamental phase in the pipeline of most of the leading optical flow estimation algorithms. The current state-of-the-art method for interpolation, EpicFlow, is a local average method based on an edge aware geodesic distance. We propose a new data-driven sparse-to-dense interpolation algorithm based on a fully convolutional network. We draw inspiration from the filling-in process in the visual cortex and introduce lateral dependencies between neurons and multi-layer supervision into our learning process. We also show the importance of the image contour to the learning process. Our method is robust and outperforms EpicFlow on competitive optical flow benchmarks with several underlying matching algorithms. This leads to state-of-the-art performance on the Sintel and KITTI 2012 benchmarks.

    Bio:
    Did my BSc In CS in the college of management. My Msc and PHD in neuroscience in the gonda center for brain research in Bar Ilan University. I spent the first half of my PhD in Proffesor Hamutal Slovin's lab investigating the visual system in the primate brain using voltage sensitive dye imaging. The second half was dedicated to deep learning and computer vision under the supervision of Professor Lior Wolf of TAU. Currently I lead the computer vision and ML field in Intuition Robotics, a company that develops a social robot designed to reduce loneliness among the elderly.

  • ceClub: Securing Internet Routing from the Ground Up

    Speaker:
    Michael Schapira (Hebrew University of Jerusalem)
    Date:
    Wednesday, 28.6.2017, 11:30
    Place:
    EE Meyer Building 861

    The Internet's communication infrastructure (TCP/IP, DNS, BGP, etc.) is alarmingly insecure, as evidenced by many high-profile incidents. I will illustrate the challenges en route to securing the Internet, and how these can be overcome, by focusing on the Internet's, arguably, biggest security hole: the vulnerability of Internet routing to traffic hijacking attacks.

    Bio:
    Michael Schapira is an associate professor at the School of Computer Science and Engineering, the Hebrew University of Jerusalem. He is also the scientific co-leader of the Fraunhofer Cybersecurity Center at Hebrew University. His research interests focus on the protocols/mechanisms that make the Internet tick (e.g., for routing, traffic management). He is interested in the design and analysis of practical (Inter)network architectures and protocols with provable guarantees (failure-resilience, optimality, security, incentive compatibility, and beyond). He is also broadly interested in the the interface of computer science, economics, and game theory (e.g., dynamic interactions in economic and computational environments, incentive-compatible computation, computational auctions, and more).

  • Probabilistic Reasoning Meets Heuristic Search

    Speaker:
    Rina Dechter - SPECIAL GUEST LECTURE - Note unusual day and time
    Date:
    Wednesday, 28.6.2017, 11:30
    Place:
    Room 337 Taub Bld.
    Affiliation:
    Donald Bren School of Information and Computer Sciences, UC Irvine
    Host:
    Benny Kimelfeld

    Graphical models, including constraint networks, Bayesian networks, Markov random fields and influence diagrams, have become a central paradigm for knowledge representation and reasoning in Artificial Intelligence, and provide powerful tools for solving problems in a variety of application domains, including coding and information theory, signal and image processing, data mining, learning, computational biology, and computer vision. Although past decades have seen considerable progress in algorithms in graphical models, many real-world problems are of such size and complexity that they remain out of reach. Advances in exact and approximate inference methods are thus crucial to address these important problems with potential impact across many computational disciplines. Exact inference is typically NP-hard, motivating the development of approximate and anytime techniques. After summarizing the main principles behind the AND/OR search guided by heuristics based on variational inference (e.g., weighted mini-bucket and cost-shifting schemes) for graphical model queries, I will focus on recent work for solving the marginal map task, a query that combines, and generalizes, optimization and summations queries and is far harder than both. The emerging solvers aim for anytime behavior that generate not only an approximation that improves with time, but also upper and lower bounds which become tighter with more time. Short Bio. ========== Rina Dechter's research centers on computational aspects of automated reasoning and knowledge representation including search, constraint processing, and probabilistic reasoning. She is a Chancellor's Professor of Computer Science at the University of California, Irvine. She holds a Ph.D. from UCLA, an M.S. degree in applied mathematics from the Weizmann Institute, and a B.S. in mathematics and statistics from the Hebrew University in Jerusalem. She is an author of Constraint Processing published by Morgan Kaufmann (2003), and Reasoning with Probabilistic and Deterministic Graphical Models: Exact Algorithms by Morgan and Claypool publishers, 2013, has co-authored close to 200 research papers, and has served on the editorial boards of: Artificial Intelligence, the Constraint Journal, Journal of Artificial Intelligence Research (JAIR), and Journal of Machine Learning Research (JMLR). She is a Fellow of the American Association of Artificial Intelligence 1994, was a Radcliffe Fellow 2005-2006, received the 2007 Association of Constraint Programming (ACP) Research Excellence Award, and she is a 2013 ACM Fellow. She has been Co-Editor- in-Chief of Artificial Intelligence since 2011. She is also co-editor with Hector Geffner and Joe Halpern of the book Heuristics, Probability and Causality: A Tribute to Judea Pearl, College Publications, 2010. ======================================== Refreshments will be served from 11:15 Lecture starts at 11:30

  • Generalizations of the Cardinality Estimation Problem and Applications to Computer Networks

    Speaker:
    Aviv Yehezkel, Ph.D. Thesis Seminar
    Date:
    Wednesday, 28.6.2017, 13:00
    Place:
    Taub 601
    Advisor:
    Prof. Reuven Cohen

    Sketch-based streaming algorithms allow efficient processing of big data. These algorithms use small fixed-size storage to store a summary ("sketch") of the input data, and use probabilistic algorithms to accurately estimate the desired quantity. A fundamental streaming problem is "cardinality estimation": given a very long stream of elements, the goal is to estimate the number of distinct elements. This is a well-known problem with numerous applications for network monitoring, security, query optimization, query execution progress, etc. This Ph.D. research focuses on several generalizations of the cardinality estimation problem. We propose new streaming algorithms, analyze their efficiency and statistical performance, and use them for solving interesting network monitoring problems. In this talk, two main results will be presented. The first is a novel estimator to the cardinality of set intersection between two sets, which is shown to outperform all previously known estimators. The second is a new framework for the cardinality estimation problem. This framework consists of a novel sketch, and a family of algorithms that use this sketch to accurately estimate the cardinality of any set expression.

  • The True Difference Between Emulation and Paravirtualization of High-Throughput I/O Devices

    Speaker:
    Arthur Kiyanovski, M.Sc. Thesis Seminar
    Date:
    Wednesday, 28.6.2017, 13:00
    Place:
    Taub 644
    Advisor:
    Prof. Dan Tsafrir

    Machine virtualization has grown in popularity in recent years with the growth of cloud computing. Virtual machines use virtual I/O devices to perform their I/O. Nowadays paravirtual I/O devices are the most popular type of virtual I/O devices due to their high performance and interposition capabilities. However paravirtual I/O devices also have disadvantages. Users need to install device drivers for paravirtual devices whenever they switch hypervisors, and hypervisor providers need to implement device drivers for all operating systems. Emulated I/O devices also allow interposition, and do not have the disadvantages of paravirtual I/O devices as they are designed to work with the device drivers of the physical devices they emulate. These device drivers come preinstalled in all major operating systems, which makes the task of switching hypervisors much easier for the users. And since the device drivers have already been written for the physical devices being emulated, hypervisor providers need not implement device drivers for emulated devices. However emulated I/O devices achieve substantially lower performance than paravirtual ones, which makes them unusable in many real world scenarios. Previous works state that the main reason for the performance difference between paravirtual and emulated I/O devices is the larger amount of exits caused by the latter. To test this claim we created a model that estimates the maximum possible throughput that can be achieved by QEMU’s emulated e1000 NIC, when taking the throughput of the paravirtual virtio-net NIC and adding the overhead of e1000’s extra exits. This model predicts a throughput difference of only 1.13X in favor of virtio-net, which is very different from the 20X throughput difference achieved in practice. This result led us to search for reasons other than exits that could explain this difference. In this work we present differences between QEMU’s virtio-net and e1000 other than exits, which we found contributing to the throughput gap between the two. For each difference we propose an improvement to e1000, inspired by virtio-net’s implementation. We then use the sidecore paradigm to reduce part of the exits caused by e1000 to further improve the throughput of e1000. We were able to reduce the throughput gap between vritio-net and e1000 down to 1.2X when the guest runs on a single core and to 1.25X on a dual core with our sidecore. Our results show that emulated I/O devices can achieve performace that is close to that of paravirtual ones, which might make emulated devices the better choice when flexibilty is more important than best performance.

  • Pixel Club: Probabilistic Pursuits on Graphs

    Speaker:
    Michael Amir (CS, Technion)
    Date:
    Thursday, 29.6.2017, 11:30
    Place:
    Taub 301

    We consider a discrete system of "ant-like" agents which pursue each other on the vertices of a graph environment. Visually reminiscent of a trail of ants, the agents emerge one by one at equal time intervals from a source vertex s and pursue each other by greedily attempting to close the distance to their immediate predecessor, the agent that emerged just before them from s, until they arrive at the destination point t. Such pursuits have been investigated before in the continuous setting and in discrete time when the underlying environment is a regular grid graph. In both these settings the agents' walks provably converge to a shortest path between s and t. Furthermore, assuming a certain natural probability distribution over the move choices of the agents on the grid (in case there are multiple shortest paths between an agent and its predecessor), the walk​​s converge to the uniform distribution over all shortest paths from s to t.

    We study, more generally, the evolution of such systems over a general graph environment G. We will exhibit surprising connections to topology, convex geometry and graph theory. For example, if every three pairwise intersecting disks of G have a non-empty (shared) intersection, the agents' walks are guaranteed to converge to the uniform distribution over all shortest paths. ​​

    *​MSc research under Prof. Alfred M. Bruckstein.

  • Sampling-on-Demand in SDN

    Speaker:
    Evgeny Moroshko, M.Sc. Thesis Seminar
    Date:
    Wednesday, 5.7.2017, 15:00
    Place:
    Taub 601
    Advisor:
    Prof. Reuven Cohen

    Sampling is an expensive network resource, because switches and routers are able to sample only a small fraction of the traffic they receive. Modern switches and routers perform uniform packet sampling, which has several important drawbacks: (i) the same flow might be unnecessarily sampled multiple times in different switches; (ii) all the flows traversing a switch whose sampling module is activated are sampled in the same rate; (iii) the sampling rate is fixed, even if the volume of the traffic changes. In this work, we propose a sampling on-demand monitoring framework. The proposed framework, presented as a component of SDN (Software Defined Network), adds a Sampling Management Module to the SDN controller. This module allows the controller to determine the sampling rate of each flow at each switch according to the monitoring goals of the network operator, while taking into account the monitoring capabilities of the switch. As part of the proposed framework, we define a new optimization problem called SAP (Sampling Allocation Problem), which has to be solved by the Sampling Management Module in order to maximize the sampling total utility. We present online and offline algorithms for solving this problem. We also present three real network management applications, executed over Mininet, which are shown to significantly benefit from the proposed framework.

  • Coding Schemes for Non-Volatile Memories

    Speaker:
    Michal Horovitz, Ph.D. Thesis Seminar
    Date:
    Thursday, 6.7.2017, 14:30
    Place:
    Taub 601
    Advisor:
    Prof. Eitan Yaakobi and Prof. Tuvi Etzion

    Flash memories is a non-volatile technology that is both electrically programmable and electrically erasable. It incorporates a set of cells maintained at a set of levels of charge to encode information. While raising the charge level of a cell is an easy operation, reducing the charge level requires the erasure of the whole block to which the cell belongs, which is a slow operation and damages the life-time of the device. I will present some of our results regarding two coding schemes which overcome the common errors in flash memories. The first scheme I will present is Write-Once Memory (WOM) codes which allow to write multiple times to the memory without decreasing the levels of the cells. The second scheme is Rank Modulation (RM) in which the information is carried by the relative ranking of the cells' charge levels and not by their absolute values. This scheme provides a more efficient programming of cells and is more robust to charge leakage. I will also mention the reconstruction problem which is come up in the DNA-storage, another non-volatile technology.

  • On Visibility and Point Clouds

    Speaker:
    Nati Kligler, M.Sc. Thesis Seminar
    Date:
    Tuesday, 11.7.2017, 11:30
    Place:
    Taub 601
    Advisor:
    Prof. A. Tal

    We introduce the concept of visibility detection within a point set to new domains. Specifically, we show that a simple representation of an image as a 3D point cloud lets us use visibility detection in classical image processing tasks, improving state-of-the-art results. Given an image, each pixel is represented as a feature point, a viewpoint is set, and points that are visible to the viewpoint are detected. What does it mean for a point to be visible? Although this question has been widely studied within computer graphics, it has never been regarded when the point set consists of feature vectors (rather than a real scene). We show that the answer to this question reveals unique information about the image, enabling us to modify state-of-the-art algorithms and improve their own results. As proof of concept, we demonstrate this idea within three applications: text image binarization, document unshadowing and stippling-style illustration.