Central limit theorems for general transportation costs

Abstract

We consider the problem of optimal transportation with general cost between an empirical measure and a general target probability on Rd , with d ≥ 1. We provide results on asymptotic stability of optimal transport potentials under minimal regularity assumptions on the costs or the underlying probability. This stability is combined with a refined linearization technique based on the sequential compactness of the closed unit ball in L2(P ) for the weak topology and the strong convergence of Cesàro means along subsequences. As a result we obtain a CLT for the transportation cost under sharp smoothness and moment assumptions, giving a positive answer to a conjecture in (Ann. Probab. 47 (2019) 926–951) for the quadratic costs.