BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York BEGIN:DAYLIGHT DTSTART:20250309T030000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:EDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20241103T010000 TZOFFSETFROM:-0400 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:EST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20241126T171920Z UID:BE067D55-B46D-440D-A7A8-B1139B4D5C4B DTSTART;TZID=America/New_York:20241114T120000 DTEND;TZID=America/New_York:20241114T130000 DESCRIPTION:This talk will highlight probabilistic computers as an emerging paradigm for domain-specific computation. Firmly connected to the widely used Markov Chain Monte Carlo algorithms widely used in physics\, statisti cs\, and ML\, the talk will show how networks of probabilistic bits\, or p -bits\, in hardware can deliver improvements in time and energy to solutio ns for ML\, optimization\, and quantum simulation.\n\nProbabilistic comput ers leverage a physics-inspired architecture with sparse connectivity and asynchronous updates\, enabling massive parallelism. Digital implementatio ns in single FPGAs show competitive performance against optimized GPUs/TPU s. Recent efforts with a distributed system of multiple FPGAs creates the “illusion” of a single\, more powerful system\, achieving near-linear speedup with minimal communication overhead.\n\nBeyond digital CMOS\, magn etic nanodevices offer intrinsic randomness\, replacing thousands of trans istors per p-bit and reducing energy per operation. Our ongoing efforts ai m to integrate these devices into energy-efficient CMOS+X systems. Compari sons with quantum computers\, GPUs/TPUs\, and coupled oscillators will ill ustrate how probabilistic computers combined with tailored algorithms coul d achieve GPU-like impact and enable new applications.\n\nSpeaker(s): \, K erem \n\n5000 Forbes Ave\, Bosch Spark Conference Room\, Scott Hall 5201\, Pittsburgh\, Pennsylvania\, United States\, 15213 LOCATION:5000 Forbes Ave\, Bosch Spark Conference Room\, Scott Hall 5201\, Pittsburgh\, Pennsylvania\, United States\, 15213 ORGANIZER:simranjs@andrew.cmu.edu SEQUENCE:16 SUMMARY:From Hardware to Algorithms: Probabilistic Computing for Machine Le arning\, Optimization\, and Quantum Simulation URL;VALUE=URI:https://events.vtools.ieee.org/m/445109 X-ALT-DESC:Description: <br /><p class="m_5359408045580785749gmail-CDt4Ke" dir="ltr">This talk will highlight probabilistic computers as an emerging paradigm for domain-specific computation. Firmly connected to the widely u sed Markov Chain Monte Carlo algorithms widely used in physics\, statistic s\, and ML\, the talk will show how networks of probabilistic bits\, or p- bits\, in hardware can deliver improvements in time and energy to solution s for ML\, optimization\, and quantum simulation.&nbsp\;</p>\n<p class="m_ 5359408045580785749gmail-CDt4Ke" dir="ltr">Probabilistic computers leverag e a physics-inspired architecture with sparse connectivity and asynchronou s updates\, enabling massive parallelism. Digital implementations in singl e FPGAs show competitive performance against optimized GPUs/TPUs. Recent e fforts with a distributed system of multiple FPGAs creates the &ldquo\;ill usion&rdquo\; of a single\, more powerful system\, achieving near-linear s peedup with minimal communication overhead.&nbsp\;</p>\n<p class="m_535940 8045580785749gmail-CDt4Ke" dir="ltr">Beyond digital CMOS\, magnetic nanode vices offer intrinsic randomness\, replacing thousands of transistors per p-bit and reducing energy per operation. Our ongoing efforts aim to integr ate these devices into energy-efficient CMOS+X systems. Comparisons with q uantum computers\, GPUs/TPUs\, and coupled oscillators will illustrate how probabilistic computers combined with tailored algorithms could achieve G PU-like impact and enable new applications.</p> END:VEVENT END:VCALENDAR