Sample complexity and weak limits of nonsmooth multimarginal Schr\"{o}dinger system with application to optimal transport barycenter

TL;DR

This paper investigates the statistical estimation and inference of entropic multimarginal optimal transport, deriving sample complexity bounds and distributional limits under weak assumptions, and introduces a new regularization method for Wasserstein barycenters.

Contribution

It provides the first sharp sample complexity bounds and distributional limits for EMOT under weak smoothness assumptions, and proposes a novel Schrödinger barycenter for optimal transport regularization.

Findings

Derived sharp sample complexity bounds for EMOT estimation.

Established distributional limits and bootstrap validity under weak smoothness.

Introduced the multimarginal Schrödinger barycenter with statistical optimality.

Abstract

Multimarginal optimal transport (MOT) has emerged as a useful framework for many applied problems. However, compared to the well-studied classical two-marginal optimal transport theory, analysis of MOT is far more challenging and remains much less developed. In this paper, we study the statistical estimation and inference problems for the entropic MOT (EMOT), whose optimal solution is characterized by the multimarginal Schr\"{o}dinger system. Assuming only boundedness of the cost function, we derive sharp sample complexity for estimating several key quantities pertaining to EMOT (cost functional and Schr\"{o}dinger coupling) from point clouds that are randomly sampled from the input marginal distributions. Moreover, with substantially weaker smoothness assumption on the cost function than the existing literature, we derive distributional limits and bootstrap validity of various key EMOT objects. As an application, we propose the multimarginal Schr\"{o}dinger barycenter as a new and natural way to regularize the exact Wasserstein barycenter and demonstrate its statistical optimality.