Schr\"odinger bridge with transport relaxation

TL;DR

This paper introduces a relaxed version of the Schr"odinger bridge problem suitable for empirical measures, providing duality, existence, uniqueness, and efficient algorithms with convergence guarantees.

Contribution

It formulates a transport-relaxed Schr"odinger bridge, derives duality, analyzes limiting behavior, and proposes convergent algorithms for practical computation.

Findings

Duality formula for the transport-relaxed bridge

Existence and uniqueness of solutions

Linear convergence of proposed algorithms

Abstract

Motivated by modern machine learning applications where we only have access to empirical measures constructed from finite samples, we relax the marginal constraints of the classical Schr\"odinger bridge problem by penalizing the transport cost between the bridge's marginals and the prescribed marginals. We derive a duality formula for this transport-relaxed bridge and demonstrate that it reduces to a finite-dimensional concave optimization problem when the prescribed marginals are discrete and the reference distribution is absolutely continuous. We establish the existence and uniqueness of solutions for both the primal and dual problems. Moreover, as the penalty blows up, we characterize the limiting bridge as the solution to a discrete Schr\"odinger bridge problem and identify a leading-order logarithmic divergence. Finally, we propose gradient ascent and Sinkhorn-type algorithms to numerically solve the transport-relaxed Schr\"odinger bridge, establishing a linear convergence rate for both algorithms.