English

Controlling Moments with Kernel Stein Discrepancies

Machine Learning 2025-06-24 v7 Machine Learning Computation

Abstract

Kernel Stein discrepancies (KSDs) measure the quality of a distributional approximation and can be computed even when the target density has an intractable normalizing constant. Notable applications include the diagnosis of approximate MCMC samplers and goodness-of-fit tests for unnormalized statistical models. The present work analyzes the convergence control properties of KSDs. We first show that standard KSDs used for weak convergence control fail to control moment convergence. To address this limitation, we next provide sufficient conditions under which alternative diffusion KSDs control both moment and weak convergence. As an immediate consequence we develop, for each q>0q > 0, the first KSDs known to exactly characterize qq-Wasserstein convergence.

Keywords

Cite

@article{arxiv.2211.05408,
  title  = {Controlling Moments with Kernel Stein Discrepancies},
  author = {Heishiro Kanagawa and Alessandro Barp and Arthur Gretton and Lester Mackey},
  journal= {arXiv preprint arXiv:2211.05408},
  year   = {2025}
}

Comments

Accepted to the Annals of Applied Probability (103 pages, 10 figures)

R2 v1 2026-06-28T05:34:51.235Z