Saved in:
Bibliographic Details
Main Authors: Glaser, Pierre, Huang, Kevin Han, Gretton, Arthur
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.13438
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908594521243648
author Glaser, Pierre
Huang, Kevin Han
Gretton, Arthur
author_facet Glaser, Pierre
Huang, Kevin Han
Gretton, Arthur
contents We perform a non-asymptotic analysis of the contrastive divergence (CD) algorithm, a training method for unnormalized models. While prior work has established that (for exponential family distributions) the CD iterates asymptotically converge at an $O(n^{-1 / 3})$ rate to the true parameter of the data distribution, we show, under some regularity assumptions, that CD can achieve the parametric rate $O(n^{-1 / 2})$. Our analysis provides results for various data batching schemes, including the fully online and minibatch ones. We additionally show that CD can be near-optimal, in the sense that its asymptotic variance is close to the Cramér-Rao lower bound.
format Preprint
id arxiv_https___arxiv_org_abs_2510_13438
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Near-Optimality of Contrastive Divergence Algorithms
Glaser, Pierre
Huang, Kevin Han
Gretton, Arthur
Machine Learning
We perform a non-asymptotic analysis of the contrastive divergence (CD) algorithm, a training method for unnormalized models. While prior work has established that (for exponential family distributions) the CD iterates asymptotically converge at an $O(n^{-1 / 3})$ rate to the true parameter of the data distribution, we show, under some regularity assumptions, that CD can achieve the parametric rate $O(n^{-1 / 2})$. Our analysis provides results for various data batching schemes, including the fully online and minibatch ones. We additionally show that CD can be near-optimal, in the sense that its asymptotic variance is close to the Cramér-Rao lower bound.
title Near-Optimality of Contrastive Divergence Algorithms
topic Machine Learning
url https://arxiv.org/abs/2510.13438