BEGIN:VCALENDAR
VERSION:2.0
CALSCALE:GREGORIAN
PRODID:UW-Madison-Physics-Events
BEGIN:VEVENT
SEQUENCE:3
UID:UW-Physics-Event-8571
DTSTART:20240221T180000Z
DTEND:20240221T190000Z
DTSTAMP:20260413T223511Z
LAST-MODIFIED:20240220T005401Z
LOCATION:Chamberlin 2104
SUMMARY:Diffusion Models - A Thermodynamic Perspective\, Physics ∩ M
 L Seminar\, Akhil Premkumar\, KICP Chicago\, University of Chicago
DESCRIPTION:Diffusion models have found immense success in modeling co
 mplex\, high dimensional data distributions. Most recently\, the text-
 to-video generation tool\, Sora by OpenAI\, uses such models to produc
 e extremely high fidelity videos from simple text prompts. In this tal
 k I will introduce a physicist-friendly intuition for diffusion models
 . Starting with first principles I will demonstrate how diffusion mode
 ls can be understood as a variational problem\, like the ones we come 
 across in physics. I will give a thermodynamic interpretation to the t
 hese models\, connecting them back to the fluctuation theorems that or
 iginally inspired their invention. Based on: arXiv:2310.04490
URL:https://www.physics.wisc.edu/events/?id=8571
END:VEVENT
END:VCALENDAR