BEGIN:VCALENDAR
VERSION:2.0
CALSCALE:GREGORIAN
PRODID:UW-Madison-Physics-Events
BEGIN:VEVENT
SEQUENCE:1
UID:UW-Physics-Event-6855
DTSTART:20220202T170000Z
DTEND:20220202T181500Z
DTSTAMP:20260414T154052Z
LAST-MODIFIED:20220130T202740Z
LOCATION:Chamberlin 5280 (Zoom link also available for online particip
 ants who signed up on our mailing list)
SUMMARY:Equivariant Neural Fields: A Roadmap Towards Generalizable Neu
 ral Representation and Inference\, Physics ∩ ML Seminar\, Ge Yang\, 
 IAIFI and MIT
DESCRIPTION:Generalization is a central problem in deep learning resea
 rch because it directly affects how much data and compute it costs to 
 achieve good performance. In other words better generalization makes b
 etter performance more accessible. In this talk\, we begin by looking 
 at a few interesting situations where modern neural networks fail to g
 eneralize. We discuss the components responsible for these failures\, 
 and ways to fix them. Then we introduce continuous neural representati
 on and neural fields as a unifying theme. As part of the roadmap\, I w
 ill lay down key technical milestones\, and specific applications in c
 ontrol\, reinforcement learning\, and scene understanding.
URL:https://www.physics.wisc.edu/events/?id=6855
END:VEVENT
END:VCALENDAR