BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:Israel BEGIN:DAYLIGHT DTSTART:20220325T030000 TZOFFSETFROM:+0200 TZOFFSETTO:+0300 RRULE:FREQ=YEARLY;BYDAY=-1FR;BYMONTH=3 TZNAME:IDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20211031T010000 TZOFFSETFROM:+0300 TZOFFSETTO:+0200 RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10 TZNAME:IST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20220325T091047Z UID:A14ED5A7-D552-422D-B4E5-97D720A4E654 DTSTART;TZID=Israel:20220322T153000 DTEND;TZID=Israel:20220322T163000 DESCRIPTION:RE: Monthly webinar - 2 Talks:\n\nTalk 1: Embracing the Resilie ncy of Deep Neural Networks: Rethinking Old Mechanisms\n\nSpeaker 1: Dr. G il Shomron\, Electrical and Computer Engineering Faculty\, Technion – Is rael Institute of Technology.\n\nTalk 2: On a Recoverability of Graph Neur al Network Representations\n\nSpeaker 1: Dr. Chaim Baskin\, Electrical and Computer Engineering Faculty\, Technion – Israel Institute of Technolog y.\n\nDate: Tuesday\, March 22nd\, 2022 3:30 PM (Israel Time)\,\n\nDear IE EE Israel chapter members\, Dear guests\,\n\nThe IEEE Computer Society Isr ael conducts a series of webinars in different areas of computer systems\, Software engineering\, Computer architectures\, data center\, accelerator s for machine learning\, security\, and more. The webinars offer insightfu l and enriching talks held by international leaders and professionals of t he computer society sector.\n\nThe next free online webinar will include t wo talks:\n\n- “Embracing the Resiliency of Deep Neural Networks: Rethin king Old Mechanisms” by Dr. Gil Shomron\, Electrical and Computer Engine ering Faculty\, Technion – Israel Institute of Technology.\n\nDeep neura l networks (DNNs) have gained tremendous momentum in recent years\, both i n academia and industry. Yet\, DNNs are compute intensive and may require billions of multiply-and-accumulate operations for a single input query. L imited resources\, such as those in IoT devices\, latency constraints\, an d high input throughput\, all drive research and development of efficient computing methods for DNN execution. In our research\, we rethink two well -known CPU methods – simultaneous multithreading (SMT) and value predict ion – and map them to the new environment introduced by DNNs\, by levera ging their unique characteristics. With SMT\, we propose a new concept of non-blocking SMT (NB-SMT)\, in which execution units are shared among seve ral computational flows to avoid idle MAC operations due to zero-valued op erands. We present and discuss the path from a data-driven “blocking” SMT design to the concept of NB-SMT\, to a fine-tuned sparsity-aware quant ization method. As for value prediction\, we present prediction schemes wh ich leverage the inherent spatial correlation in CNN feature maps to predi ct zero-valued activations. By speculating which activations will be zero- valued\, we potentially reduce the required MAC operations.\n\n- “On a R ecoverability of Graph Neural Network Representations” by Dr. Chaim Bask in\, Electrical and Computer Engineering Faculty\, Technion – Israel Ins titute of Technology.\n\nDespite their growing popularity\, graph neural n etworks (GNNs) still have multiple unsolved problems\, including finding m ore expressive aggregation methods\, propagation of information to distant nodes\, and training on large-scale graphs. Understanding and solving suc h problems require developing analytic tools and techniques. In this work\ , we propose the notion of recoverability\, which is tightly related to in formation aggregation in GNNs\, and based on this concept\, develop the me thod for GNN embedding analysis. We define recoverability theoretically an d propose a method for its efficient empirical estimation. We demonstrate\ , through extensive experimental results on various datasets and different GNN architectures\, that estimated recoverability correlates with aggrega tion method expressivity and graph sparsification quality. Therefore\, we believe that the proposed method could provide an essential tool for under standing the roots of the aforementioned problems\, and potentially lead t o a GNN design that overcomes them.\n\nThe Webinar is free\, but pre-regis tration is required. So\, please sign up using the below link https://tech nion.zoom.us/webinar/register/WN_-vxCyWMpRO2sSVyJT2h40A and the Zoom sessi on details will be provided after registration.\n\nPlease contact us for a ny further details and updates on the series of IEEE Computer Society Webi nars.\n\nWe are looking forward to your participation and future collabora tion.\n\nProf. Avi Mendelson Prof. Freddy Gabbay\nAvi.mendelson@technion.a c.il freddyg@ruppin.ac.il\n\nChairman Vice-Chair\n\nVirtual: https://event s.vtools.ieee.org/m/307733 LOCATION:Virtual: https://events.vtools.ieee.org/m/307733 ORGANIZER:freddyg@ruppin.ac.il SEQUENCE:1 SUMMARY:Monthly webinar: "Embracing the Resiliency of Deep Neural Networks: Rethinking Old Mechanisms" and "On a Recoverability of Graph Neural Netwo rk Representations" URL;VALUE=URI:https://events.vtools.ieee.org/m/307733 X-ALT-DESC:Description: <br /><p>RE: Monthly webinar - 2 Talks:</p>\n<p>Tal k 1: <strong>Embracing the Resiliency of Deep Neural Networks: Rethinking Old Mechanisms</strong></p>\n<p>Speaker 1: <strong>Dr. Gil Shomron\, </str ong><strong>Electrical and Computer Engineering Faculty\, Technion &ndash\ ; Israel Institute of Technology.</strong></p>\n<p><strong>&nbsp\;</strong ></p>\n<p>Talk 2: <strong>On a Recoverability of Graph Neural Network Repr esentations</strong></p>\n<p>Speaker 1: <strong>Dr. Chaim Baskin\, </stron g><strong>Electrical and Computer Engineering Faculty\, Technion &ndash\; Israel Institute of Technology.</strong></p>\n<p><strong>&nbsp\;</strong>< /p>\n<p>Date: <strong>Tuesday\, </strong><strong>March 22<sup>nd</sup>\, 2 022</strong><strong> 3:30 PM (Israel Time)</strong><strong>\,</strong></p> \n<p>&nbsp\;</p>\n<p>Dear IEEE Israel chapter members\,&nbsp\;Dear guests\ ,</p>\n<p>&nbsp\;</p>\n<p>The IEEE Computer Society Israel conducts a seri es of webinars in different areas of computer systems\, Software engineeri ng\, Computer architectures\, data center\, accelerators for machine learn ing\, security\, and more. The webinars offer insightful and enriching tal ks held by international leaders and professionals of the computer society sector.</p>\n<p>&nbsp\;</p>\n<p>The next free online webinar will include two talks:</p>\n<ol>\n<li>&ldquo\;Embracing the Resiliency of Deep Neural Networks: Rethinking Old Mechanisms&rdquo\; by Dr. Gil Shomron\, Electric al and Computer Engineering Faculty\, Technion &ndash\; Israel Institute o f Technology.</li>\n</ol>\n<p>&nbsp\;</p>\n<p>Deep neural networks (DNNs) have gained tremendous momentum in recent years\, both in academia and ind ustry. Yet\, DNNs are compute intensive and may require billions of multip ly-and-accumulate operations for a single input query. Limited resources\, such as those in IoT devices\, latency constraints\, and high input throu ghput\, all drive research and development of efficient computing methods for DNN execution. In our research\, we rethink two well-known CPU methods &ndash\; simultaneous multithreading (SMT) and value prediction &ndash\; and map them to the new environment introduced by DNNs\, by leveraging the ir unique characteristics. With SMT\, we propose a new concept of non-bloc king SMT (NB-SMT)\, in which execution units are shared among several comp utational flows to avoid idle MAC operations due to zero-valued operands. We present and discuss the path from a data-driven &ldquo\;blocking&rdquo\ ; SMT design to the concept of NB-SMT\, to a fine-tuned sparsity-aware qua ntization method. As for value prediction\, we present prediction schemes which leverage the inherent spatial correlation in CNN feature maps to pre dict zero-valued activations. By speculating which activations will be zer o-valued\, we potentially reduce the required MAC operations.</p>\n<p>&nbs p\;</p>\n<p>&nbsp\;</p>\n<ol start="2">\n<li>&ldquo\;On a Recoverability o f Graph Neural Network Representations&rdquo\; by Dr. Chaim Baskin\, Elect rical and Computer Engineering Faculty\, Technion &ndash\; Israel Institut e of Technology.</li>\n</ol>\n<p>Despite their growing popularity\, graph neural networks (GNNs) still have multiple unsolved problems\, including f inding more expressive aggregation methods\, propagation of information to distant nodes\, and training on large-scale graphs. Understanding and sol ving such problems require developing analytic tools and techniques. In th is work\, we propose the notion of recoverability\, which is tightly relat ed to information aggregation in GNNs\, and based on this concept\, develo p the method for GNN embedding analysis. We define recoverability theoreti cally and propose a method for its efficient empirical estimation. We demo nstrate\, through extensive experimental results on various datasets and d ifferent GNN architectures\, that estimated recoverability correlates with aggregation method expressivity and graph sparsification quality. Therefo re\, we believe that the proposed method could provide an essential tool f or understanding the roots of the aforementioned problems\, and potentiall y lead to a GNN design that overcomes them.</p>\n<p>&nbsp\;</p>\n<p>&nbsp\ ;</p>\n<p>The Webinar is free\, but pre-registration is required. So\, ple ase sign up using the below link <a href="https://technion.zoom.us/webinar /register/WN_-vxCyWMpRO2sSVyJT2h40A">https://technion.zoom.us/webinar/regi ster/WN_-vxCyWMpRO2sSVyJT2h40A</a> and the Zoom session details will be pr ovided after registration.</p>\n<p>&nbsp\;</p>\n<p>Please contact us for a ny further details and updates on the series of IEEE Computer Society Webi nars.</p>\n<p>&nbsp\;</p>\n<p>We are looking forward to your participation and future collaboration.</p>\n<p>&nbsp\;</p>\n<p>Prof. Avi Mendelson&nbs p\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;& nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp \;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&n bsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\; &nbsp\;&nbsp \;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\; P rof. Freddy Gabbay<br /><a href="mailto:Avi.mendelson@technion.ac.il">Avi. mendelson@technion.ac.il</a> &nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nb sp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\; &nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbs p\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\; <a href="mailto:fredd yg@ruppin.ac.il">freddyg@ruppin.ac.il</a></p>\n<p>Chairman &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; & nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nb sp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp \; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\;Vice-Chair</p>\n<p>&nbsp \;</p> END:VEVENT END:VCALENDAR