BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:www.tcs.tifr.res.in/event/1544
DTSTAMP:20250411T051043Z
SUMMARY:Sampling is as easy as learning the score: theory for diffusion mod
 els with minimal data assumptions
DESCRIPTION:Speaker: Agniv Bandyopadhyay (TIFR)\n\nAbstract: \nWe will cons
 ider the problem of proving theoretical convergence guarantees for score-b
 ased generative models (SGMs)\, such as DDPMs\, which are foundational to 
 large-scale generative systems like DALL·E 2. We will prove that\, given 
 L2-accurate score estimates\, SGMs can efficiently sample from a broad cla
 ss of realistic data distributions without relying on restrictive assumpti
 ons like log-concavity or log-Sobolev-ness. The convergence rate scales po
 lynomially with problem parameters and matches the best-known complexity b
 ounds for Langevin diffusion discretization. Proving this convergence rate
  relies on a clever application of Girsanov's theorem\, which is a celebra
 ted theorem in stochastic calculus\, which we will explore during this tal
 k. \nThis talk will be based on results presented in the paper: Chen\, S
 .\, Chewi\, S.\, Li\, J.\, Li\, Y.\, Salim\, A.\, & Zhang\, A. R. (2022). 
 Sampling is as easy as learning the score: theory for diffusion models wit
 h minimal data assumptions. (https://arxiv.org/abs/2209.11215)\n
URL:https://www.tcs.tifr.res.in/web/events/1544
DTSTART;TZID=Asia/Kolkata:20250411T170000
DTEND;TZID=Asia/Kolkata:20250411T180000
LOCATION:A-201 (STCS Seminar Room)
END:VEVENT
END:VCALENDAR