MMODELYST
Papers/Wasserstein Contraction of Coordinate Ascent Variational Inference
PAP

Wasserstein Contraction of Coordinate Ascent Variational Inference

May 28, 2026

arXiv
Abstract

We study the contraction in Wasserstein distance of the coordinate ascent variational inference algorithm. This is shown to hold under a transport-information inequality at the fixed points and a functional smoothness condition. The results are general and sharp, allow for local convergence guarantees, hold for general smooth manifolds, and also in some non-smooth spaces. We consider applications to Bayesian Gaussian Mixture Models, and high-dimensional Bayesian Probit Regression, and Logistic Regression with Pólya-Gamma random variables (i.e. Jaakkola-Jordan's algorithm).

Select text to highlight · click a highlight to remove · saved in this browser only
Authors
Rocco Caprio, Adrien Corenflos, Sam Power
Your notes (browser-local)
saved
arXiv:2605.30253