Convergence of the Sinkhorn algorithm when the Schr\"odinger problem has no solution

TL;DR

This paper investigates the behavior of the Sinkhorn algorithm in degenerate cases where the Schr"odinger problem has no solution, revealing its limit points and their use in relaxed problem formulations, with applications in cell biology.

Contribution

It characterizes the Sinkhorn algorithm's convergence behavior in degenerate cases and introduces a method to compute solutions of a relaxed Schr"odinger problem.

Findings

The algorithm alternates between two limit points when no solution exists.

Limit points can be used to solve a relaxed Schr"odinger problem.

The support of the relaxed solution has a predictable shape.

Abstract

The Sinkhorn algorithm is the most popular method for solving the entropy minimization problem called the Schr\"odinger problem: in the non-degenerate cases, the latter admits a unique solution towards which the algorithm converges linearly. Here, motivated by recent applications of the Schr\"odinger problem with respect to structured stochastic processes (such as increasing ones), we study the Sinkhorn algorithm in degenerate cases where it might happen that no solution exist at all. We show that in this case, the algorithm ultimately alternates between two limit points. Moreover, these limit points can be used to compute the solution of a relaxed version of the Schr\"odinger problem, which appears as the $\Gamma$-limit of a problem where the marginal constraints are replaced by asymptotically large marginal penalizations, exactly in the spirit of the so-called unbalanced optimal transport. Finally, our work focuses on the support of the solution of the relaxed problem, giving its typical shape and designing a procedure to compute it quickly. We showcase promising numerical applications related to a model used in cell biology.