Center for Advanced Intelligence Project

Date and Time:
May 31, 2024: 10:00 am – 11:00 am (JST)
Venue: Online and Open Space at the RIKEN AIP Nihonbashi office*
*The Open Space; only available to AIP researchers.

Title:
Variational inference via Wasserstein gradient flows

Speaker:
Sinho Chewi (Institute for Advanced Study)

Abstract:
Variational inference (VI), which seeks to approximate the Bayesian posterior by a more tractable distribution within a variational family, has been widely advocated as a scalable alternative to MCMC. However, obtaining non-asymptotic convergence guarantees has been a longstanding challenge. In this talk, I will argue that viewing this problem as optimization over the Wasserstein space of probability measures equipped with the optimal transport metric leads to the design of principled algorithms which exhibit strong practical performance and are backed by rigorous theory. In particular, we address Gaussian VI, as well as (non-parametric) mean-field VI.

Bio:
Sinho Chewi is currently a postdoctoral researcher at the Institute for Advanced Study, and he will soon join Yale University as an assistant professor in the Statistics and Data Science department. Previously, he received his PhD at the Massachusetts Institute of Technology under the supervision of Professor Philippe Rigollet.