BEGIN:VCALENDAR
VERSION:2.0
CALSCALE:GREGORIAN
PRODID:UW-Madison-Physics-Events
BEGIN:VEVENT
SEQUENCE:0
UID:UW-Physics-Event-6924
DTSTART:20220420T160000Z
DTEND:20220420T171500Z
DTSTAMP:20260414T152806Z
LAST-MODIFIED:20220415T011922Z
LOCATION:Online Seminar: Please sign up for our mailing list at www.ph
 ysicsmeetsml.org for zoom link. We will also livestream the talk in Ch
 amberlin 5280.
SUMMARY:Effective Theory of Deep Neural Networks\, Physics ∩ ML Semi
 nar\, Sho Yaida\, Meta AI
DESCRIPTION:Large neural networks perform extremely well in practice\,
  providing the backbone of modern machine learning. The goal of this t
 alk is to provide a blueprint for theoretically analyzing these large 
 models from first principles. In particular\, we’ll overview how the
  statistics and dynamics of deep neural networks drastically simplify 
 at large width and become analytically tractable. In so doing\, we’l
 l see that the idealized infinite-width limit is too simple to capture
  several important aspects of deep learning such as representation lea
 rning. To address them\, we’ll step beyond the idealized limit and s
 ystematically incorporate finite-width corrections.
URL:https://www.physics.wisc.edu/events/?id=6924
END:VEVENT
END:VCALENDAR