{ "id": "2304.04451", "version": "v1", "published": "2023-04-10T08:35:44.000Z", "updated": "2023-04-10T08:35:44.000Z", "title": "Quantitative contraction rates for Sinkhorn algorithm: beyond bounded costs and compact marginals", "authors": [ "Giovanni Conforti", "Alain Durmus", "Giacomo Greco" ], "comment": "51 pages", "categories": [ "math.PR", "math.OC" ], "abstract": "We show non-asymptotic geometric convergence of Sinkhorn iterates to the Schr\\\"odinger potentials, solutions of the quadratic Entropic Optimal Transport problem on $\\mathbb{R}^d$. Our results hold under mild assumptions on the marginal inputs: in particular, we only assume that they admit an asymptotically positive log-concavity profile, covering as special cases log-concave distributions and bounded smooth perturbations of quadratic potentials. More precisely, we provide exponential $\\mathrm{L}^1$ and pointwise convergence of the iterates (resp. their gradient and Hessian) to Schr\\\"odinger potentials (resp. their gradient and Hessian) for large enough values of the regularization parameter. As a corollary, we establish exponential convergence of Sinkhorn plans and bridges w.r.t. a symmetric relative entropy. Up to the authors' knowledge, these are the first results which establish geometric convergence of Sinkhorn algorithm in a general setting without assuming bounded cost functions or compactly supported marginals. Our results are proven following a probabilistic approach that rests on integrated semiconvexity estimates for Sinkhorn iterates that are of independent interest.", "revisions": [ { "version": "v1", "updated": "2023-04-10T08:35:44.000Z" } ], "analyses": { "subjects": [ "49Q22", "90C25", "49N05", "93E20", "47D07" ], "keywords": [ "quantitative contraction rates", "sinkhorn algorithm", "bounded cost", "compact marginals", "quadratic entropic optimal transport problem" ], "note": { "typesetting": "TeX", "pages": 51, "language": "en", "license": "arXiv", "status": "editable" } } }