Variational inference via Wasserstein gradient flows

MaRDI QIDQ6400705arXivFDO

Authors Marc Lambert, Sinho Chewi, Francis Bach, Silvère Bonnabel, Philippe Rigollet

Publication date 31 May 2022

Published in arXiv

Has companion code repository https://github.com/marc-h-lambert/w-vi

arXiv classification

cs.LGmath.STstat.MLstat.TH

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 $pi$, VI aims at producing a simple but effective approximation $hatpi$ to $pi$ 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 work, we propose principled methods for VI, in which $hatpi$ 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 $pi$ is log-concave.

Cited in

(1)

• Nesterov acceleration in optimizing over probability measures

This page was built for publication: Variational inference via Wasserstein gradient flows

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6400705)