English

A gradient descent perspective on Sinkhorn

Optimization and Control 2020-06-11 v3

Abstract

We present a new perspective on the popular Sinkhorn algorithm, showing that it can be seen as a Bregman gradient descent (mirror descent) of a relative entropy (Kullback-Leibler divergence). This viewpoint implies a new sublinear convergence rate with a robust constant.

Cite

@article{arxiv.2002.03758,
  title  = {A gradient descent perspective on Sinkhorn},
  author = {Flavien Léger},
  journal= {arXiv preprint arXiv:2002.03758},
  year   = {2020}
}
R2 v1 2026-06-23T13:36:44.050Z