BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:To Intrinsic Dimension and Beyond: Efficient Sampling in Diffusion
  Models - Dr. Yuting Wei\, Associate Professor in the Statistics and Data 
 Science Department at the Wharton School\, University of Pennsylvania
DTSTART:20251111T140000Z
DTEND:20251111T150000Z
UID:TALK240517@talks.cam.ac.uk
CONTACT:Fernando Ruiz Mazo
DESCRIPTION:*Abstract*: The denoising diffusion probabilistic model (DDPM)
  has become a cornerstone of generative AI. While sharp convergence guaran
 tees have been established for DDPM\, the iteration complexity typically s
 cales with the ambient data dimension of target distributions\, leading to
  overly conservative theory that fails to explain its practical efficiency
 . This has sparked recent efforts to understand how DDPM can achieve sampl
 ing speed-ups through automatic exploitation of intrinsic low dimensionali
 ty of data. This talk explores two key scenarios: (1) For a broad class of
  data distributions with intrinsic dimension k\, we prove that the iterati
 on complexity of the DDPM scales nearly linearly with k\, which is optimal
  under the KL divergence metric\; (2) For mixtures of Gaussian distributio
 ns with k components\, we show that DDPM learns the distribution with iter
 ation complexity that grows only logarithmically in k. These results provi
 de theoretical justification for the practical efficiency of diffusion mod
 els.\n\n*Bio*: Dr. Yuting Wei is an Associate Professor in the Statistics 
 and Data Science Department at the Wharton School\, University of Pennsylv
 ania. Prior to that\, Dr. Wei spent two years at Carnegie Mellon Universit
 y as an assistant professor and one year at Stanford University as a Stein
 's Fellow. She received her Ph.D. in statistics at the University of Calif
 ornia\, Berkeley. She was the recipient of the 2025 Gottfried E. Noether E
 arly Career Scholar Award\, Google Research Scholar Award\, NSF Career awa
 rd\, and the Erich L. Lehmann Citation from the Berkeley statistics depart
 ment. Her research interests include high-dimensional and non-parametric s
 tatistics\, reinforcement learning\, and diffusion models.\n\n*This talk i
 s co-hosted by the Computer Laboratory AI Research Group and the Informed-
 AI Hub.*\n
LOCATION:Lecture Theatre 2\, Computer Laboratory\, William Gates Building
END:VEVENT
END:VCALENDAR