Using Sinkhorn in the JKO scheme adds linear diffusion

2025/02/18 by Baradat, Aymeric, Hraivoronska, Anastasiia, Santambrogio, Filippo · 2 citations
#Analysis of PDEs (math.AP) #FOS: Mathematics

paper · doi:10.48550/arxiv.2502.12666

Abstract

The JKO scheme is a time-discrete scheme of implicit Euler type that allows to construct weak solutions of evolution PDEs which have a Wasserstein gradient structure. The purpose of this work is to study the effect of replacing the classical quadratic optimal transport problem by the Schrödinger problem (a.k.a. the entropic regularization of optimal transport, efficiently computed by the Sinkhorn algorithm) at each step of this scheme. We find that if ε is the regularization parameter of the Schrödinger problem, and τ is the time step parameter, considering the limit τ,ε→ 0 with \fracετ → α∈ ℝ₊ results in adding the term \fracα2 Δρ on the right-hand side of the limiting PDE. In the case α= 0 we improve a previous result by Carlier, Duval, Peyré and Schmitzer (2017).

Cited by

• Gradient Flows of Potential Energies in the Geometry of Sinkhorn Divergences
• A synthetic approach to comparison principles for variational problems, with applications to optimal transport

Related

• Gradient Flows of Potential Energies in the Geometry of Sinkhorn Divergences
• Convergence of the Sinkhorn algorithm when the Schrödinger problem has no solution
• On the total variation Wasserstein gradient flow and the TV-JKO scheme
• Wasserstein Mirror Gradient Flow as the limit of the Sinkhorn Algorithm
• Stability of Schrödinger Potentials and Convergence of Sinkhorn's Algorithm