Saved in:
Bibliographic Details
Main Authors: Li, Lei, Wang, Yuliang
Format: Preprint
Published: 2022
Subjects:
Online Access:https://arxiv.org/abs/2207.09304
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913745211490304
author Li, Lei
Wang, Yuliang
author_facet Li, Lei
Wang, Yuliang
contents We establish a sharp uniform-in-time error estimate for the Stochastic Gradient Langevin Dynamics (SGLD), which is a widely-used sampling algorithm. Under mild assumptions, we obtain a uniform-in-time $O(η^2)$ bound for the KL-divergence between the SGLD iteration and the Langevin diffusion, where $η$ is the step size (or learning rate). Our analysis is also valid for varying step sizes. Consequently, we are able to derive an $O(η)$ bound for the distance between the invariant measures of the SGLD iteration and the Langevin diffusion, in terms of Wasserstein or total variation distances. Our result can be viewed as a significant improvement compared with existing analysis for SGLD in related literature.
format Preprint
id arxiv_https___arxiv_org_abs_2207_09304
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle A sharp uniform-in-time error estimate for Stochastic Gradient Langevin Dynamics
Li, Lei
Wang, Yuliang
Probability
Machine Learning
65C20, 68Q25, 60H30
We establish a sharp uniform-in-time error estimate for the Stochastic Gradient Langevin Dynamics (SGLD), which is a widely-used sampling algorithm. Under mild assumptions, we obtain a uniform-in-time $O(η^2)$ bound for the KL-divergence between the SGLD iteration and the Langevin diffusion, where $η$ is the step size (or learning rate). Our analysis is also valid for varying step sizes. Consequently, we are able to derive an $O(η)$ bound for the distance between the invariant measures of the SGLD iteration and the Langevin diffusion, in terms of Wasserstein or total variation distances. Our result can be viewed as a significant improvement compared with existing analysis for SGLD in related literature.
title A sharp uniform-in-time error estimate for Stochastic Gradient Langevin Dynamics
topic Probability
Machine Learning
65C20, 68Q25, 60H30
url https://arxiv.org/abs/2207.09304