Colloquia and Seminars

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

Upcoming Colloquia & Seminars

  • CGGC Seminar: Designing N-PolyVector Fields with Complex Polynomials

    Speaker:
    Amir Vaxman (Vienna University of Technology)
    Date:
    Sunday, 8.3.2015, 13:30
    Place:
    Room 337-8 Taub Bld.

    N-PolyVector fields are introduced, which are a generalization of N-RoSy fields for which the vectors are neither necessarily orthogonal nor rotationally symmetric. A novel representation for N-PolyVectors as the root sets of complex polynomials is given, in addition to the analysis of their topological and geometric properties. A smooth N-PolyVector field can be efficiently generated by solving a sparse linear system without integer variables. This flexibility of N-PolyVector fields can be explored for the design of conjugate vector fields, offering an intuitive tool to generate planar quadrilateral meshes. Further extensions to curl-free fields and other applications will be discussed.

  • Pixel Club: Ricci flow for Image Processing and Computer Vision

    Speaker:
    Ezri Sonn (EE, Technion)
    Date:
    Tuesday, 10.3.2015, 11:30
    Place:
    EE Meyer Building 1061

    A discrete model of the Ricci flow for images is presented, using a purely combinatorial method for calculating the Ricci curvature, based on R. Forman's work on generalized Laplacians for cell complexes. The adaptation of Forman's curvature function for cell complexes to images is natural and straightforward, using geometric properties of the image as a surface embedded in Euclidean space. It is shown that the framework presented is far more applicable to images than other existing models. This work focuses on analyzing the proposed scheme and the different methods of its implementation. Several weighting methods for the calculation of the Ricci curvature are suggested and compared, and different implementations of the discrete Ricci flow are explored. The model is analyzed in terms of curvature validity, conformality, convergence and numerical stability. Simulation results show that the discrete Ricci flow has a powerful effect on the image and its curvature, and evolves the surface metric in a way that is consistent with theoretical expectations. Finally, some applications for image processing and computer vision are proposed, such as forward-and-backward Ricci flow, high dynamic range imaging, and change detection in aerial images.

  • Online Learning and Competitive Analysis: a Unified Approach

    Speaker:
    Shahar Chen, Ph.D. Thesis Seminar
    Date:
    Tuesday, 10.3.2015, 12:30
    Place:
    Taub 601
    Advisor:
    Prof. Seffi Naor and Dr. Niv Buchbinder

    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 research we provide several contributions in this direction. First, we provide a single unified algorithm which by parameter tuning, interpolates between optimal regret for learning from experts (in online learning) and optimal competitive ratio for the metrical task systems problem (MTS) (in competitive analysis), improving on the results of Blum and Burch (1997). The algorithm also allows us to obtain new regret bounds against "drifting" experts, which might be of independent interest. Moreover, our approach allows us to go beyond experts/MTS, obtaining similar unifying results for structured action sets and "combinatorial experts", whenever the setting has a certain matroid structure. A complementary direction of our research tries to "borrow" various learning techniques, specifically focusing on the online convex optimization domain, in order to obtain new results in the competitive analysis framework. We show how \emph{regularization}, a fundamental method in machine learning and particularly in the field of online learning, can be applied to obtain new results in the area of competitive analysis. We also show how \emph{convex conjugacy} and \emph{Fenchel duality}, other powerful techniques used in online convex optimization and learning, can be used in the competitive analysis setting, allowing us to cope with a richer world of online optimization problems.

  • CGGC Seminar: Persistent Homology

    Speaker:
    Alexandre Djerbetian (CS, Technion)
    Date:
    Sunday, 22.3.2015, 13:30
    Place:
    Room 337-8 Taub Bld.

    Shapes are usually described by their geometrical characteristics. Obviously, a rabbit doesn't look like a horse, because of its curvature for instance. However, some characteristics of a shape are independent of its geometry. A rugby ball is still a ball, and a doughnut looks like a tire. Those are called topological features of the shape, and homology is one the tool that mathematicians use to separate a sphere from a torus. A rabbit being equivalent to a horse, this often makes topology be a neglected tool in computer science for obvious reasons. However, some topological knowledge can sometimes help us have a better understanding of the underlying shape, behind the geometry, and then help make meaningful topological simplification. In this talk, I will introduce the concept and the algorithm of persistent homology, a tool to classify the importance of the topological features of arbitrary shapes. We will for instance count the number of mountains and valleys in a given geography, and sort them by topological importance.

  • ceClub: Variability SmartBalance: A Sensing-Driven Linux Load Balancer for Energy Efficiency of Heterogeneous MPSoCs

    Speaker:
    Prof. Alex Nicolau (University of California, Irvine)
    Date:
    Wednesday, 25.3.2015, 11:30
    Place:
    Taub 6

    A short overview of the NSF Variability Expedition will be given, followed by an overview of a particular result: SmartBalance.

    Due to increased demand for higher performance and better energy efficiency, MPSoCs are deploying heterogeneous architectures with architecturally differentiated core types. However, the traditional Linux-based operating system is unable to exploit this heterogeneity since existing kernel load balancing and scheduling approaches lack support for aggressively heterogeneous architectural configurations (e.g. beyond two core types). In this paper we present SmartBalance: a sensing-driven closed-loop load balancer for aggressively heterogeneous MPSoCs that performs load balancing using a sense-predict-balance paradigm. SmartBalance can efficiently manage the chip resources while opportunistically exploiting the workload variations and performance-power trade-offs of different core types. When compared to the standard vanilla Linux kernel load balancer, our per-thread and per-core performance-power-aware scheme shows an improvement in energy efficiency (throughput/Watt) of over 50% for benchmarks from the PARSEC benchmark suite executing on a heterogeneous MPSoC with 4 different core types and over 20% w.r.t. state-of-the-art ARM's global task scheduling (GTS) scheme for octa-core big.Little architecture.

    Bio: Alex Nicolau received his Ph.D. in Computer Science from Yale University in 1984, and served on the faculty of the computer science department at Cornell University until 1988. That year he joined the University of California, Irvine as an associate professor, where he serves as full professor since 1992 and department chair since 2013.

    The author of over 300 conference and journal articles and many books, Alex chaired numerous international conferences (e.g., ACM International Supercomputing Conference (ICS) and Principles and Practice of Parallel Programming) and is editor in chief of the International Journal of Parallel Programming, the oldest journal in that field. He is also an IEEE Fellow.

  • ceClub: Deep learning with NVIDIA GPUs

    Speaker:
    Jonathan Cohen (NVIDIA Corporation)
    Date:
    Thursday, 26.3.2015, 13:30
    Place:
    Taub 6

    NVIDIA GPUs are powering a revolution in machine learning. With the rise of deep learning algorithms, in particular deep convolutional neural networks, computers are learning to see, hear, and understand the world around us in ways never before possible. Image recognition and detection systems are getting close to and in some cases surpassing human-level performance. I will talk about deep learning in the context of several new NVIDIA initiatives ranging from hardware platforms, software tools and libraries, and our recently announced DRIVE PX module for autonomous driving.

    Bio:
    Jonathan Cohen is Director of Engineering for NVIDIA's GPU-accelerated deep learning software platform. Before moving to the product side of NVIDIA, Mr. Cohen spent three years as a senior research scientist with NVIDIA Research developing scientific computing and real-time physical simulation applications on NVIDIA's massively parallel GPUs.

  • CS M.Sc. & Ph.D. Alumni Event

    CS M.Sc. & Ph.D. Alumni Event

    Date:
    Thursday, 2.4.2015, 14:30
    Place:
    CS Taub

    We are very excited to invite you to the MSc & PhD alumni event. The event will be held on Thursday, April 2nd at the CS faculty. A detailed invitation will be followed.

    If you are familiar with other MSc & PhD alumni - we'd love to get their contact details to make sure everyone received the invitation.

    Hope to see you,