English

Wasserstein Gradient Flows of MMD Functionals with Distance Kernels under Sobolev Regularization

Analysis of PDEs 2025-05-06 v2

Abstract

We consider Wasserstein gradient flows of maximum mean discrepancy (MMD) functionals MMDK2(,ν)\text{MMD}_K^2(\cdot, \nu) for positive and negative distance kernels K(x,y):=±xyK(x,y) := \pm |x-y| and given target measures ν\nu on R\mathbb{R}. Since in one dimension the Wasserstein space can be isometrically embedded into the cone C(0,1)L2(0,1)\mathcal C(0,1) \subset L_2(0,1) of quantile functions, Wasserstein gradient flows can be characterized by the solution of an associated Cauchy problem on L2(0,1)L_2(0,1). While for the negative kernel, the MMD functional is geodesically convex, this is not the case for the positive kernel, which needs to be handled to ensure the existence of the flow. We propose to add a regularizing Sobolev term H1(0,1)2|\cdot|^2_{H^1(0,1)} corresponding to the Laplacian with Neumann boundary conditions to the Cauchy problem of quantile functions. Indeed, this ensures the existence of a generalized minimizing movement for the positive kernel. Furthermore, for the negative kernel, we demonstrate by numerical examples how the Laplacian rectifies a "dissipation-of-mass" defect of the MMD gradient flow.

Keywords

Cite

@article{arxiv.2411.09848,
  title  = {Wasserstein Gradient Flows of MMD Functionals with Distance Kernels under Sobolev Regularization},
  author = {Richard Duong and Nicolaj Rux and Viktor Stein and Gabriele Steidl},
  journal= {arXiv preprint arXiv:2411.09848},
  year   = {2025}
}

Comments

29 pages, 13 figures. Corrected mistake in functional in Lemma 3.1. Added Remark 3.9 on long-time asymptotics. Added clarifications and references throughout the paper

R2 v1 2026-06-28T20:00:36.559Z