Convergence rate of general entropic optimal transport costs
We investigate the convergence rate of the optimal entropic cost to the optimal transport cost as the noise parameter . We show that for a large class of cost functions on (for which optimal plans are not necessarily unique or induced by a transport map) and compactly supported and marginals, one has . Upper bounds are obtained by a block approximation strategy and an integral variant of Alexandrov's theorem. Under an infinitesimal twist condition on , i.e. invertibility of , we get the lower bound by establishing a quadratic detachment of the duality gap in dimensions thanks to Minty's trick.