קולוקוויום וסמינרים

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


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

Academic Calendar at Technion site.

קולוקוויום וסמינרים בקרוב

  • Pixel Club: Joint Design of Optics and Post-Processing Algorithms Based on Deep Learning for Generating Advanced Imaging Features

    דובר:
    שי אלמלם (אונ' תל-אביב(
    תאריך:
    יום שלישי, 25.2.2020, 11:30
    מקום:
    חדר 1061, בניין מאייר, הפקולטה להנדסת חשמל

    The recent and on-going Deep-Learning (DL) revolution, introduces a paradigm shift in almost all disciplines of signal processing. Traditionally, Computer Vision (CV) and Image Processing (IP) methods were based on hand-crafted feature extraction from the initial optical image, and then some hand-crafted classifier/filter was defined to achieve the desired result IP/CV. The machine learning approach seeks to learn the 'classifier' stage using data (either labeled or unlabeled), i.e. the decision rule is not explicitly derived a priori from the data, but implicitly optimized using a large set of examples. DL methods take this approach to the next level, so that the feature extraction stage is also learned. In such an approach the design is done by defining a parameterized computational model, and then training it (i.e. optimizing its parameters) end-to-end, using the data. After the tremendous success of DL for IP/CV applications, these days almost every signal processing task is analyzed using such tools. In the presented work, the DL design revolution is brought one step deeper, into the optical image formation process. By considering the lens as an analog signal processor of the incoming optical wavefront (originating from the scene), the optics is modeled as an additional 'layer' in a DL model, and its parameters are optimized jointly with the 'conventional' DL layers, end-to-end. This design scheme allows the introduction of unique feature encoding in the intermediate optical image, since the lens 'has access' to information that is lost in conventional 2D imaging. Therefore, such design allows a holistic design of the entire IP/CV system. The proposed design approach will be presented with several applications: an extended Depth-Of-Field (DOF) camera; a passive depth estimation solution based on a single image from a single camera; non-uniform motion deblurring; and enhanced stereo camera with extended dynamic range and self-calibration abilities. Experimental results will be presented and discussed.

    Short bio:
    Shay Elmalem is a Ph.D. candidate at the Faculty of Engineering, Tel-Aviv University, under the supervision of Dr. Raja Giryes (until recently also under the supervision of the late Prof. Emanuel Marom). His research interests include computational imaging, with applications to optical design, image processing, and computer vision.

  • Learning-based design of ultrasound imaging systems

    דובר:
    סנקט ודולה, הרצאה סמינריונית למגיסטר
    תאריך:
    יום רביעי, 26.2.2020, 11:00
    מקום:
    טאוב 012 (מרכז רב תכליתי)
    מנחה:
    Prof. Alex Bronstein and Dr. Michael Zibulevsky

    Medical ultrasound is a widespread imaging modality owing its popularity to cost efficiency, portability, speed, and lack of harmful ionizing radiation. In this work, we demonstrate that replacing the traditional ultrasound processing pipeline with a data-driven, learnable counterpart leads to significant improvement in image quality. Moreover, we demonstrate that greater improvement can be achieved through a learning-based design of the transmitted beam patterns simultaneously with learning an image reconstruction pipeline. We evaluate our method on an in-vivo first-harmonic cardiac ultrasound dataset acquired from volunteers and demonstrate the significance of the learned pipeline and transmit beam patterns on the image quality when compared to standard transmit and receive beamformers used in high frame-rate US imaging. We believe that the presented methodology provides a fundamentally different perspective on the classical problem of ultrasound beam pattern design.

  • Formal Program Repair

    דובר:
    בת-חן רוטנברג, הרצאה סמינריונית לדוקטורט
    תאריך:
    יום רביעי, 26.2.2020, 11:30
    מקום:
    טאוב 601
    מנחה:
    Prof. Orna Grumberg

    The manual detection, examination and repair of computer bugs are all notoriously difficult tasks that programmers face daily. Automated program repair receives a program with a bug, and outputs a set of changes to the program that would make it bug-free. We focus on formal program repair, which means that programs are repaired with respect to a formal specification. Formal program repair, in contrast to test-based repair, guarantees that programs meet the specification for all inputs, and not just a selected set. In this talk, we portray the challenges posed by the formal repair problem, and discuss our approach for overcoming them. We present several algorithms developed as part of this research, which make use of mathematical and logical tools from the world of program verification, such as constraint solvers and software model checkers. We have implemented these algorithms and have tested them for the repair of buggy student submissions to programming assignments. Our results show that our algorithms are capable of producing effective repairs for these submissions, very efficiently.

  • Learning feasible and efficient MR imaging

    דובר:
    תומר וייס, הרצאה סמינריונית למגיסטר
    תאריך:
    יום רביעי, 26.2.2020, 11:45
    מקום:
    טאוב 012 (מרכז רב תכליתי)
    מנחה:
    Prof. A. Bronstein

    Magnetic Resonance Imaging (MRI) has long been considered to be among "the gold standards" of diagnostic medical imaging. The long acquisition times, however, render MRI prone to motion artifacts, let alone their adverse contribution to the relative high costs of MRI examination. Over the last few decades, multiple studies have focused on the development of both physical and post-processing methods for accelerated acquisition of MRI scans. These two approaches, however, have so far been addressed separately. On the other hand, recent works in optical computational imaging have demonstrated growing success of concurrent learning-based design of data acquisition and image reconstruction schemes. In this work, we propose a novel approach to the learning of optimal schemes for conjoint acquisition and reconstruction of MRI scans, with the optimization carried out simultaneously with respect to the time-efficiency of data acquisition and the quality of resulting reconstructions. To be of a practical value, the schemes are encoded in the form of general k-space trajectories, whose associated magnetic gradients are constrained to obey a set of predefined hardware requirements (as defined in terms of, e.g., peak currents and maximum slew rates of magnetic gradients). With this proviso in mind, we propose a novel algorithm for the end-to-end training of a combined acquisition-reconstruction pipeline using a deep neural network with differentiable forward- and back-propagation operators. We also demonstrate the effectiveness of the proposed solution in application to both image reconstruction and image segmentation, reporting substantial improvements in terms of acceleration factors as well as the quality of these end tasks.

  • Breaking the Bluetooth Pairing the Fixed Coordinate Invalid Curve Attack

    דובר:
    ליאור נוימן, הרצאה סמינריונית למגיסטר
    תאריך:
    יום רביעי, 4.3.2020, 11:00
    מקום:
    טאוב 601
    מנחה:
    Prof. Eli Biham

    Bluetooth is a widely deployed standard for wireless communications between mobile devices. It uses authenticated Elliptic Curve Diffie-Hellman for its key exchange. In this paper we show that the authentication provided by the Bluetooth pairing protocols is insufficient and does not provide the promised MitM protection. We present a new attack that modifies the y-coordinates of the public keys (while preserving the x-coordinates). The attack compromises the encryption keys of all of the current Bluetooth authenticated pairing protocols, provided both paired devices are vulnerable. Specifically, it successfully compromises the encryption keys of 50% of the Bluetooth pairing attempts, while in the other 50% the pairing of the victims is terminated. The affected vendors have been informed and patched their products accordingly, and the Bluetooth specification had been modified to address the new attack. We named our new attack the ``Fixed Coordinate Invalid Curve Attack''. Unlike the well known ``Invalid Curve Attack'' of Biehl et. al. which recovers the private key by sending multiple specially crafted points to the victim, our attack is a MitM attack which modifies the public keys in a way that lets the attacker deduce the shared secret.

  • Leveraging Machine Learning Algorithms in Online Portfolio Selection

    דובר:
    גיא עוזיאל, הרצאה סמינריונית לדוקטורט
    תאריך:
    יום רביעי, 4.3.2020, 13:30
    מקום:
    טאוב 601
    מנחה:
    Prof. R. El-Yaniv

    Online Portfolio Selection, aiming to optimize the allocation of wealth across a set of assets, is a fundamental research problem in computational finance and machine learning. Despite the theoretical challenges, the implementation of a real-world trading system is extremely challenging. This issue has been extensively studied across several research communities, including finance, statistics, coding and information theory and machine learning. This long-standing problem, however, still poses many challenges. We begin with a discussion of the problem of online portfolio selection in the presence of transaction costs, and we present two novel algorithms, enabling a trader to enhance any existing commission-oblivious algorithms. Our experiments which were conducted on common benchmarks show that the two new algorithms achieve state-of-the-art results. We then present an approach to handle the risk incurred while trading. First, we review multi-objective online learning, where we propose a novel framework and algorithm to address this problem in case the underlying process is stationary and ergodic. We prove that under mild conditions our algorithm is universal and thus asymptotically achieves the best possible outcome in hindsight. Later on, we show how this method can be utilized to incorporate the well-known risk proxy, conditional value at risk (CVaR) in online portfolio selection. Finally, we deal with the pattern matching algorithms and propose a novel approach to incorporate the learning of a suitable kernel using a deep neural network, in an online manner. The talk summarizes work presented in 5 papers, 4 of which were published/accepted to NIPS, AISTATS.

  • Hypernetworks and a New Feedback Model

    דובר:
    Lior Wolf - COLLOQUIUM LECTURE
    תאריך:
    יום שלישי, 31.3.2020, 14:30
    מקום:
    חדר 337 טאוב.
    השתייכות:
    School of Computer Science, Tel-Aviv University
    מארח:
    Yuval Filmus

    Hypernetworks, also known as dynamic networks, are neural networks in which the weights of at least some of the layers vary dynamically based on the input. Such networks have composite architectures in which one network predicts the weights of another network. I will briefly describe the early days of dynamic layers and present recent results from diverse domains: 3D reconstruction from a single image, image retouching, electrical circuit design, decoding block codes, graph hypernetworks for bioinformatics, and action recognition in video. Finally, I will present a new hypernetwork-based model for the role of feedback in neural computations. Short Bio: ========== Lior Wolf is a faculty member at the School of Computer Science at Tel Aviv University and a research scientist at Facebook AI Research. Before, he was a postdoc working with Prof. Poggio at CBCL, MIT. He received his PhD working with Prof. Shashua at the Hebrew U, Jerusalem. ====================================== Refreshments will be served from 14:15 Lecture starts at 14:30

  • תאריך:
    מקום:
  • Second-order Optimization for Machine Learning, Made Practical

    דובר:
    Tomer Koren - COLLOQUIUM LECTURE
    תאריך:
    יום שלישי, 5.5.2020, 14:30
    מקום:
    חדר 337 טאוב.
    השתייכות:
    School of Computer Science at Tel Aviv University
    מארח:
    Yuval Filmus

    Optimization in machine learning, both theoretical and applied, is presently dominated by first-order gradient methods such as stochastic gradient descent. Second-order optimization methods---that involve second-order derivatives and/or second-order statistics of the data---have become far less prevalent despite strong theoretical properties, due to their impractical computation, memory and communication costs. I will present some recent theoretical, algorithmic and infrastructural advances that allow for overcoming these challenges in using second-order methods and obtaining significant performance gains in practice, at very large scale, and on highly-parallel computing architectures. Short Bio: =========== Tomer Koren is an Assistant Professor in the School of Computer Science at Tel Aviv University since Fall 2019. Previously, he was a Senior Research Scientist at Google Brain, Mountain View. He received his PhD in December 2016 from the Technion - Israel Institute of Technology, where his advisor was Prof. Elad Hazan. His research interests are in machine learning and optimization. ===================================== Rereshments will be served from 14:15 Lecture starts at 14:30