BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/Chicago BEGIN:DAYLIGHT DTSTART:20260308T030000 TZOFFSETFROM:-0600 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:CDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20251102T010000 TZOFFSETFROM:-0500 TZOFFSETTO:-0600 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:CST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20251128T013528Z UID:944E3A1C-FCA0-4649-BEE5-D5A33328CB04 DTSTART;TZID=America/Chicago:20251121T110000 DTEND;TZID=America/Chicago:20251121T120000 DESCRIPTION:Title: Understanding and using the embedding spaces of large ge nerative models\n\nSpeaker: Anand Sarwate\, Professor of ECE\, Rutgers Uni versity\n\nAbstract:\n\nTraining massive ML/AI models on massive amounts o f data supposedly creates latent representations or features that are “u niversal” in the sense that the large model acts as a feature extractor that maps inputs into an embedding space. In this talk I will discuss rece nt empirical work that looks at embeddings for generative models. In parti cular\, I will describe an approach that uses a third model as a “micros cope” to uncover differences between two other models. Simple methods on the embedding space of the “microscope” model show that outputs of di fferent models are distinguishable\, which potentially opens the door to s everal applications. Time permitting\, I will describe other insights abou t embedding spaces.\n\nBio:\n\nAnand D. Sarwate is currently a Professor o f Electrical and Computer Engineering at Rutgers University. Prior to join ing Rutgers he was a Research Assistant Professor at TTI-Chicago and a pos tdoc at the ITA Center at UCSD. He received undergraduate degrees in mathe matics and electrical engineering from MIT (2002) and a PhD from UC Berkel ey (2008). His research interests include information theory\, machine lea rning\, signal processing\, optimization\, and privacy and security. Dr. S arwate is a Distinguished Lecturer of the IEEE Information Theory Society for 2024--2025 and is on the Board of Governors of the IEEE Information Th eory Society.\n\nLocation:\n\nUniversity of Illinois Chicago\, Lecture Cen ter C4\n\nSpeaker(s): Anand Sarwate\, \n\nBldg: Lecture Center C4\, Univer sity of Illinois Chicago East Campus\, Chicago\, Illinois\, United States\ , 60607 LOCATION:Bldg: Lecture Center C4\, University of Illinois Chicago East Camp us\, Chicago\, Illinois\, United States\, 60607 ORGANIZER:devroye@uic.edu SEQUENCE:29 SUMMARY:IEEE Information Theory Society Distinguished Lecture: Understandin g and using the embedding spaces of large generative models URL;VALUE=URI:https://events.vtools.ieee.org/m/513227 X-ALT-DESC:Description: <br /><p>Title: Understanding and using the embeddi ng spaces of large generative models</p>\n<p>Speaker: Anand Sarwate\, Prof essor of ECE\, Rutgers University</p>\n<p>Abstract:&nbsp\;</p>\n<p>Trainin g massive ML/AI models on massive amounts of data supposedly creates laten t representations or features that are &ldquo\;universal&rdquo\; in the se nse that the large model acts as a feature extractor that maps inputs into an embedding space. In this talk I will discuss recent empirical work tha t looks at embeddings for generative models. In particular\, I will descri be an approach that uses a third model as a &ldquo\;microscope&rdquo\; to uncover differences between two other models. Simple methods on the embedd ing space of the &ldquo\;microscope&rdquo\; model show that outputs of dif ferent models are distinguishable\, which potentially opens the door to se veral applications. Time permitting\, I will describe other insights about embedding spaces.<br><br>Bio:</p>\n<p>Anand D. Sarwate is currently a Pro fessor of Electrical and Computer Engineering at Rutgers University. Prior to joining Rutgers he was a Research Assistant Professor at TTI-Chicago a nd a postdoc at the ITA Center at UCSD. He received undergraduate degrees in mathematics and electrical engineering from MIT (2002) and a PhD from U C Berkeley (2008). His research interests include information theory\, mac hine learning\, signal processing\, optimization\, and privacy and securit y. Dr. Sarwate is a Distinguished Lecturer of the IEEE Information Theory Society for 2024--2025 and is on the Board of Governors of the IEEE Inform ation Theory Society.</p>\n<p>Location:</p>\n<p>University of Illinois Chi cago\, Lecture Center C4</p> END:VEVENT END:VCALENDAR