Saved in:
Bibliographic Details
Main Authors: Li, Lei, Liu, Jian-Guo, Wang, Yuliang
Format: Preprint
Published: 2023
Subjects:
Online Access:https://arxiv.org/abs/2301.06769
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929471754338304
author Li, Lei
Liu, Jian-Guo
Wang, Yuliang
author_facet Li, Lei
Liu, Jian-Guo
Wang, Yuliang
contents We consider the geometric ergodicity of the Stochastic Gradient Langevin Dynamics (SGLD) algorithm under nonconvexity settings. Via the technique of reflection coupling, we prove the Wasserstein contraction of SGLD when the target distribution is log-concave only outside some compact set. The time discretization and the minibatch in SGLD introduce several difficulties when applying the reflection coupling, which are addressed by a series of careful estimates of conditional expectations. As a direct corollary, the SGLD with constant step size has an invariant distribution and we are able to obtain its geometric ergodicity in terms of $W_1$ distance. The generalization to non-gradient drifts is also included.
format Preprint
id arxiv_https___arxiv_org_abs_2301_06769
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Geometric ergodicity of SGLD via reflection coupling
Li, Lei
Liu, Jian-Guo
Wang, Yuliang
Probability
Machine Learning
60H35, 65C40, 37A25
We consider the geometric ergodicity of the Stochastic Gradient Langevin Dynamics (SGLD) algorithm under nonconvexity settings. Via the technique of reflection coupling, we prove the Wasserstein contraction of SGLD when the target distribution is log-concave only outside some compact set. The time discretization and the minibatch in SGLD introduce several difficulties when applying the reflection coupling, which are addressed by a series of careful estimates of conditional expectations. As a direct corollary, the SGLD with constant step size has an invariant distribution and we are able to obtain its geometric ergodicity in terms of $W_1$ distance. The generalization to non-gradient drifts is also included.
title Geometric ergodicity of SGLD via reflection coupling
topic Probability
Machine Learning
60H35, 65C40, 37A25
url https://arxiv.org/abs/2301.06769