BIMSA > Seminar on Control Theory and Nonlinear Filtering Variational inference via Wasserstein gradient flows
Variational inference via Wasserstein gradient flows
Organizer
Shing Toung Yau
Speaker
Jia Yi Kang
Time
Tuesday, December 20, 2022 9:00 PM - 9:30 PM
Venue
Online
Abstract
Along with Markov chain Monte Carlo (MCMC) methods, variational inference (VI) has emerged as a central computational approach to large-scale Bayesian inference. Rather than sampling from the true posterior π, VI aims at producing a simple but effective approximation πˆ to π for which summary statistics are easy to compute. However, unlike the well-studied MCMC methodology, algorithmic guarantees for VI are still relatively less well-understood. In this report, I will introduce a paper in which authors propose principled methods for VI, in which πˆ is taken to be a Gaussian or a mixture of Gaussians, which rest upon the theory of gradient flows on the Bures–Wasserstein space of Gaussian measures. Akin to MCMC, it comes with strong theoretical guarantees when π is log-concave.