Ricci curvature and $W_1$-exponential convergence of Markov processes on graphs
Lingyan Cheng, Ruinan Li, Liming Wu

TL;DR
This paper characterizes Ricci curvature lower bounds for continuous-time Markov processes on graphs using optimal couplings, and establishes explicit exponential convergence rates with applications to Glauber dynamics.
Contribution
It extends Ollivier's Ricci curvature results from discrete to continuous time Markov processes on graphs and provides explicit convergence estimates.
Findings
Ricci curvature lower bounds relate to exponential convergence rates.
Constructed optimal coupling generator for continuous-time processes.
Applied results to Glauber dynamics under Dobrushin conditions.
Abstract
In this paper, we show that the Ricci curvature lower bound in Ollivier's Wasserstein metric sense of a continuous time jumping Markov process on a graph can be characterized by some optimal coupling generator and provide the construction of this latter. Some previous results of Ollivier for discrete time Markov chains are generalized to the actual continuous time case. We propose a comparison technique with some death-birth process on to obtain some explicit exponential convergence rate, by modifying the metric. A counterpart of Zhong-Yang's estimate is established in the case where the Ricci curvature with repsect to the graph metric is nonnegative. Moreover we show that the Lyapunov function method for the exponential convergence works with some explicit quantitative estimates, once if the Ricci curvature is bounded from below by a negative constant. Finally we present…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsGeometric Analysis and Curvature Flows · Point processes and geometric inequalities · Geometry and complex manifolds
