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

  • אונ' תל-אביב

    דובר:
    תום טירר
    תאריך:
    יום שלישי, 26.2.2019, 11:30
    מקום:
    חדר 337, בניין טאוב למדעי המחשב

    Inverse problems appear in many applications, such as image deblurring, inpainting and super-resolution. The common approach to address them is to design a specific algorithm (or recently - a deep neural network) for each problem. The Plug-and-Play (P&P) framework, which has been recently introduced, allows solving general inverse problems by leveraging the impressive capabilities of existing denoising algorithms. While this fresh strategy has found many applications, a burdensome parameter tuning is often required in order to obtain high-quality results. In this work, we propose an alternative method for solving inverse problems using off-the-shelf denoisers, which requires less parameter tuning (can be also translated into less pre-trained denoising neural networks). First, we transform a typical cost function, composed of fidelity and prior terms, into a closely related, novel optimization problem. Then, we propose an efficient minimization scheme with a plug-and-play property, i.e., the prior term is handled solely by a denoising operation. Finally, we present an automatic tuning mechanism to set the method’s parameters. We provide a theoretical analysis of the method, and empirically demonstrate its impressive results for image inpainting, deblurring and super-resolution. For the latter, we also present an image-adaptive learning approach that further improves the results.

    *PhD student under supervision of Prof. Raja Giryes.

  • Lazy Evaluation Methods for Complex Event Processing

    דובר:
    איליה קולצינסקי, הרצאה סמינריונית לדוקטורט
    תאריך:
    יום שלישי, 26.2.2019, 12:30
    מקום:
    טאוב 601
    מנחה:
    Prof. A. Schuster

    Rapid advances in data-driven applications over recent years have intensified the need for efficient mechanisms capable of real-time monitoring and detecting arbitrarily complex patterns in massive data streams. Complex event processing (CEP) is a prominent technology widely employed for performing this task in many areas, including online finance, security monitoring, credit card fraud detection, and IoT (Internet of Things) technologies. An increasingly active and rapidly developing area of academic research, CEP functionality is also provided by multiple open source libraries and commercial data analysis platforms. CEP engines operate by collecting basic data items arriving from input data streams and using them to infer complex events according to the patterns defined by the system users. To that end, data items are combined into higher-level entities matching the pattern-specified structure. In order to guarantee detection correctness, a CEP system is required to actively maintain all subsets of data items that might eventually become a part of a successful pattern match. As a result, the overall number of such potential matches grows exponentially with the size and the sophistication of the pattern being detected. Considering that real-life patterns typically incorporate a highly convoluted structure and may consist of 10 or more events connected by increasingly complex operators and predicates, this situation introduces a crucial performance bottleneck, making complex event processing virtually infeasible even for large businesses capable of acquiring extensive computation power. In addition, modern CEP applications are often required to process hundreds or even thousands of patterns and streams in parallel under tight real-time constraints, which increases the magnitude of the problem. In this talk, we present a novel solution to overcome the exponential resource requirements of complex event processing. Our solution is based on the principle of 'statistic-based lazy evaluation'. Under this paradygm, incoming data items are allowed to be processed in an order different from their natural order of appearance in an input stream. As a result, statistical properties of the underlying data, such as the occurrence rates of different types of items and the selectivities (probabilities of success) of the predicates, can be utilized to generate efficient evaluation plans providing close-to-optimal detection performance. We devise and describe an efficient lazy evaluation mechanism for complex event processing based on nondeterministic finite automata. By exploiting the similarity of our problem to the well-known problem of join query plan generation, we develop a procedure for adapting the existing join-based algorithms to the CEP domain, thus creating a new family of algorithms for generating practically efficient pattern detection plans. As the statistical data properties required for plan generation are rarely known in advance and may change dynamically, we present an efficient and precise mechanism that continuously estimates the current statistic values of the required data characteristics on-the-fly and, if and whenever necessary, adapts the evaluation plan accordingly. Finally, we extend our methods to multi-pattern CEP environment, demonstrating how the lazy evaluation approach can facilitate common subexpression sharing between different patterns in a workload. An extensive theoretical and empirical analysis of our innovations demonstrates their superiority over state-of-the-art approaches.

  • GAIA: An OS Page Cache for Heterogeneous Systems

    דובר:
    טניה ברוכמן, הרצאה סמינריונית למגיסטר
    תאריך:
    יום רביעי, 27.2.2019, 11:30
    מקום:
    Electrical Eng. Building 861
    מנחה:
    Prof. M. Silberstein

    We propose a principled approach to integrating GPU memory with an OS page cache. We design GAIA, a weakly-consistent page cache that spans CPU and GPU memories. GAIA enables the standard mmap system call to map files into the GPU address space, thereby enabling data-dependent GPU accesses to large files and efficient write-sharing between the CPU and GPUs. Under the hood, GAIA (1) integrates lazy release consistency among physical memories into the OS page cache while maintaining backward compatibility with CPU processes and unmodified GPU kernels; (2) improves CPU I/O performance by using data cached in GPU memory, and (3) optimizes readahead prefetcher to support accesses to caches in GPUs. We prototype GAIA in Linux and evaluate it on NVIDIA Pascal GPUs. We show up to 3 speedup in CPU file I/O and up to 8 in unmodified realistic workloads such as Gunrock GPU-accelerated graph processing, image collage, and microscopy i

  • Advanced Geometric Methods in Machining and Additive Manufacturing

    דובר:
    בן עזר, הרצאה סמינריונית לדוקטורט
    תאריך:
    יום ראשון, 3.3.2019, 13:30
    מקום:
    טאוב 337
    מנחה:
    Prof. G. Elber

    In this thesis, we show how geometry, and specifically parametric freeform geometry, can be used to solve manufacturing related problems. This thesis deals with both additive manufacturing (specifically 3D printing) and the more traditional (subtractive manufacturing) machining process. The first topic handled is the representation and manufacturing of functionally graded material (FGM) objects in 3D printing. Specifically, this thesis presents the use of parametric (trimmed) volumes to manufacture FGM objects. Using the underlying parametrization offered by parametric volumes enables the succinct specification of material composition in the printer's native resolution, without resorting to discrete volumetric representations (voxels) and their associated problems. The related chapter also presents several FGM objects (which would require billions of voxels to represent) fabricated using our methods. The second topic related to additive manufacturing is the use of non-planar print-paths or layers. In traditional 3D printing, objects are decomposed into a set of thin, essentially 2D, planar layers. These layers are manufactured one after the other, which greatly simplifies the planning of the 3D printing process. However, by using non-planar layouts, objects with better material properties can be manufactured. This thesis shows how non-planar covering print-paths can be generated, and how they can be 3D printed to fabricate objects. This work fully supports general objects, not limited to some very specific class of geometries, that can be fabricated using non-planar print-paths, and presents several examples of such fabricated objects. The methods presented here can also be applied to scenarios involving non-planar fiber layouts, toward composites, in 3D printing and to hybrid manufacturing. With regard to machining, this thesis presents algorithms that enable the generation of collision free 5-axis tool-paths for convex machining tools. Unlike previous efforts, the methods presented here rely on bounding sufficiently small surface patches rather than sampling of the machined geometry, with all the deficiencies that sampling can entail. By bounding the position, normal, and normal curvature properties of freeform surface patches not only is the tool-path assured to be globally collision free, it can remain so even after it is optimized using continuous optimization methods. The introduced patch subdivision and bounding procedures also enable the use of configuration space methods to help navigate between the surface patches.

  • יום עיון בסייבר ואבטחת מידע 2019

    CYBERDAY 2019

    תאריך:
    יום שלישי, 12.3.2019, 09:30
    מקום:
    Technion

    יום עיון בסייבר ואבטחת מידע 2019

    יום העיון בסייבר ואבטחת מידע 2019 באירגונם של פרופ' אלי ביהם, פרופ' שרה ביתן ומרכז המחקר לאבטחת סייבר ע"ש הירושי פוג'יאוורה, יתקיים ביום שלישי, 12 במרס 2019 בטכניון.

    מרצים אורחים:


    Prof. Bart Preneel (KU Leuven, Belgium)
    Prof. Adi Shamir (The Weizmann Institute of Science)


    רוב ההרצאות יינתנו בעברית ובהפסקה תתקיים תצוגת פוסטרים: מוזמנים סטודנטים לתארים מתקדמים וכל המעוניין להציג בה את מחקרם ולהבטיח את מקומם מראש בהקדם האפשרי בכתובת: cyber at technion.ac.il עד 28 בפברואר, לפי הפרטים באתר הכנס.

    פרטים נוספים, תוכנית מלאה (תעודכן בקרוב) והרשמה (חופשית אך נדרשת).

    מצפים לראותכם!

  • תאריך:
    מקום:
  • From Cognitive Biases to the Communication Complexity of Local Search

    דובר:
    Shahar Dobzinski - COLLOQUIUM LECTURE
    תאריך:
    יום שלישי, 2.4.2019, 14:30
    מקום:
    חדר 337 טאוב.
    השתייכות:
    Weizmann Institute, Applied Math and Computer Science Dept.
    מארח:
    Yuval Filmus

    In this talk I will tell you how analyzing economic markets where agents have cognitive biases has led to better understanding of the communication complexity of local search procedures. We begin the talk with studying combinatorial auctions with bidders that exhibit endowment effect. In most of the previous work on cognitive biases in algorithmic game theory (e.g., [Kleinberg and Oren, EC'14] and its follow-ups) the focus was on analyzing the implications and mitigating their negative consequences. In contrast, we show how cognitive biases can sometimes be harnessed to improve the outcome. Specifically, we study Walrasian equilibria in combinatorial markets. It is well known that a Walrasian equilibrium exists only in limited settings, e.g., when all valuations are gross substitutes, but fails to exist in more general settings, e.g., when the valuations are submodular. We consider combinatorial settings in which bidders exhibit the endowment effect, that is, their value for items increases with ownership. Our main result here shows that when the valuations are submodular even a mild level of endowment effect suffices to guarantee the existence of Walrasian equilibrium. In fact, we show that in contrast to Walrasian equilibria with standard utility maximizers bidders -- in which the equilibrium allocation must be a global optimum -- when bidders exhibit endowment effect any local optimum can be an equilibrium allocation. This raises the natural question of understanding the complexity of computing a local maximum in combinatorial markets. We reduce it to the following communication variant of local search: there is some fixed, commonly known graph G. Alice holds f_A and Bob holds f_B, both are functions that specify a value for each vertex. The goal is to find a local maximum of f_A+f_B, i.e., a vertex v for which f_A(v)+f_B(v) >= f_A(u)+f_B(u) for every neighbor u of v. We prove that finding a local maximum requires polynomial (in the number of vertices) bits of communication. Based on joint works with Moshe Babaioff, Yakov Babichenko, Noam Nisan, and Sigal Oren.

  • Some systems engineering problems and a little bit of theory

    דובר:
    Eitan Bachmat - COLLOQUIUM LECTURE
    תאריך:
    יום שלישי, 16.4.2019, 14:30
    מקום:
    חדר 337 טאוב.
    השתייכות:
    Ben-Gurion University
    מארח:
    Yuval Filmus

    We will consider the SITA server farm scheduling policies which were introduced and studied by Harchol-Balter and her collaborators. In particular we will discuss a recent INFOCOM 17 paper of Doncel, Aalto and Ayesta and relate one of the tables there to the work of Riemann on the functional equation of the zeta function. We will also consider certain disk drive scheduling problems (travelling salesman on a disk drive) and then relate it space-time geometry and in particular gravitational lensing for positive mass particles. Finally we will relate, via airplane boarding, the disk scheduling problem with server farm scheduling. Short Bio: ========== Eitan Bachmat received his PhD in mathematics from M.I.T. and is currently a professor of computer science at Ben-Gurion University. Eitan works on problems in the areas of Storage Systems, Performance analysis, Systems engineering, Operations research, Autism, Personalized Medicine and Digital healthcare, using techniques from Physics, Mathematics and Computer Science. He is an expert on airplane boarding and express line queues and how they relate to relativity theory and number theory respectively. Eitan also works closely with industry and various organizations on diverse applications and holds 38 patents. ========================== Refreshments will be served from 14:15 Lecture starts at 14:30

  •  יום פתוח לתארים מתקדמים במדעי המחשב

    CS Open Day For Graduate Studies

    תאריך:
    יום רביעי, 17.4.2019, 12:15
    מקום:
    חדר 337 טאוב.

    היום הפתוח לקראת ההרשמה לשנה"ל תש"פ מזמין בוגרי תואר ראשון מצטיינים מכל האוניברסיטאות להגיע לטכניון ולהתרשם מהפקולטות למדעי המחשב, לפגוש חברי סגל וסטודנטים לתארים מתקדמים ולשמוע הרצאה מרתקת מפי דר' טל כהן, דירקטור פיתוח תוכנה, גוגל: "האם תואר מתקדם אכן מקדם?"
    .
    האירוע יתקיים ביום ד', 17 באפריל 2019, בין השעות 12:15-15:00, בבניין טאוב למדעי המחשב, חדר 337 (קומה 3).

    תוכנית היום תכלול סקירה על הלימודים ותנאי הקבלה וכן הרצאות מדעיות. לכל מועמד תתאפשר פגישה אישית עם סגן הדיקן לתארים מתקדמים בפקולטה.

    המעוניינים להגיע לאירוע מתבקשים להירשם מראש.

    פרטים נוספים ותוכנית מלאה.

  • Computer Agents that Interact Proficiently with People

    דובר:
    Sarit Kraus - COLLOQUIUM LECTURE
    תאריך:
    יום שלישי, 30.4.2019, 14:30
    מקום:
    חדר 337 טאוב.
    השתייכות:
    Bar-Ilan University
    מארח:
    Yuval Filmus

    Automated agents that interact proficiently with people can be useful in supporting, training or replacing people in complex tasks. The inclusion of people presents novel problems for the design of automated agents’ strategies. People do not necessarily adhere to the optimal, monolithic strategies that can be derived analytically. Their behavior is affected by a multitude of social and psychological factors. In this talk I will show how combining machine learning techniques for human modelling, human behavioral models, formal decision-making and game theory approaches enables agents to interact well with people. Applications include intelligent agents that help drivers reduce energy consumption, agents that support rehabilitation, employer-employee negotiation and agents that support a human operator in managing a team of low-cost mobile robots in search and rescue tasks. Short Bio: =========== Sarit Kraus (Ph.D. Computer Science, Hebrew University, 1989) is a Professor of Computer Science at Bar-Ilan University. Her research is focused on intelligent agents and multi-agent systems (including people and robots). Kraus was awarded the IJCAI Computers and Thought Award, ACM SIGART Agents Research award, the EMET prize and was twice the winner of the IFAAMAS influential paper award. She is AAAI, ECCAI and ACM fellow and a recipient of the advanced ERC grant.

  • On Optimization and Generalization in Deep Learning

    דובר:
    Amir Globerson - COLLOQUIUM LECTURE
    תאריך:
    יום שלישי, 4.6.2019, 14:30
    מקום:
    חדר 337 טאוב.
    השתייכות:
    Tel-Aviv University
    מארח:
    Yuval Filmus

    TBA