Convergence Rates for Regularized Optimal Transport via Quantization
Optimization and Control
2023-06-22 v3 Probability
Machine Learning
Abstract
We study the convergence of divergence-regularized optimal transport as the regularization parameter vanishes. Sharp rates for general divergences including relative entropy or regularization, general transport costs and multi-marginal problems are obtained. A novel methodology using quantization and martingale couplings is suitable for non-compact marginals and achieves, in particular, the sharp leading-order term of entropically regularized 2-Wasserstein distance for all marginals with finite -moment.
Cite
@article{arxiv.2208.14391,
title = {Convergence Rates for Regularized Optimal Transport via Quantization},
author = {Stephan Eckstein and Marcel Nutz},
journal= {arXiv preprint arXiv:2208.14391},
year = {2023}
}
Comments
Forthcoming in 'Mathematics of Operations Research'