English

Sharp $L^q$-Convergence Rate in $p$-Wasserstein Distance for Empirical Measures of Diffusion Processes

Probability 2024-08-20 v1

Abstract

For a class of (non-symmetric) diffusion processes on a length space, which in particular include the (reflecting) diffusion processes on a connected compact Riemannian manifold, the exact convergence rate is derived for (E[Wpq(μT,μ)])1q(T)({\mathbb E} [{\mathbb W}_p^q(\mu_T,\mu)])^{\frac{1}{q}} (T \to \infty) uniformly in (p,q)[1,)×(0,)(p,q)\in [1,\infty) \times (0,\infty), where μT\mu_T is the empirical measure of the diffusion process, μ\mu is the unique invariant probability measure, and Wp{\mathbb W}_p is the pp-Wasserstein distance. Moreover, when the dimension parameter is less than 44, we prove that ETW22(μT,μ)Ξ(T)q0{\mathbb E} |T {\mathbb W}_2^2(\mu_T,\mu)-\Xi(T)|^q \to 0 as TT\to\infty for any q1q\ge 1, where Ξ(T)\Xi(T) is explicitly given by eigenvalues and eigenfunctions for the symmetric part of the generator.

Keywords

Cite

@article{arxiv.2408.09116,
  title  = {Sharp $L^q$-Convergence Rate in $p$-Wasserstein Distance for Empirical Measures of Diffusion Processes},
  author = {Feng-Yu Wang and Bingyao Wu and Jie-Xiang Zhu},
  journal= {arXiv preprint arXiv:2408.09116},
  year   = {2024}
}