English

Wasserstein Diffusion on Multidimensional Spaces

Probability 2024-04-25 v3 Functional Analysis Metric Geometry

Abstract

Given any closed Riemannian manifold MM, we construct a reversible diffusion process on the space P(M){\mathcal P}(M) of probability measures on MM that is (i) reversible w.r.t.~the entropic measure Pβ{\mathbb P}^\beta on P(M){\mathcal P}(M), heuristically given as dPβ(μ)=1ZeβEnt(μm) dP(μ);d\mathbb{P}^\beta(\mu)=\frac{1}{Z} e^{-\beta \, \text{Ent}(\mu| m)}\ d\mathbb{P}^*(\mu); (ii) associated with a regular Dirichlet form with carr\'e du champ derived from the Wasserstein gradient in the sense of Otto calculus EW(f)=lim infgf 12P(M)Wg2(μ) dPβ(μ);{\mathcal E}_W(f)=\liminf_{g\to f}\ \frac12\int_{{\mathcal P}(M)} \big\|\nabla_W g\big\|^2(\mu)\ d{\mathbb P}^\beta(\mu); (iii) non-degenerate, at least in the case of the nn-sphere and the nn-torus.

Keywords

Cite

@article{arxiv.2401.12721,
  title  = {Wasserstein Diffusion on Multidimensional Spaces},
  author = {Karl-Theodor Sturm},
  journal= {arXiv preprint arXiv:2401.12721},
  year   = {2024}
}

Comments

New result on Large Deviation Principle (Thm. 3.9 + 3.10); corrected proof for Lemma 2.6

R2 v1 2026-06-28T14:24:39.670Z