English

Sparsity of Quadratically Regularized Optimal Transport: Bounds on concentration and bias

Optimization and Control 2024-10-07 v1 Probability

Abstract

We study the quadratically regularized optimal transport (QOT) problem for quadratic cost and compactly supported marginals μ\mu and ν\nu. It has been empirically observed that the optimal coupling πϵ\pi_\epsilon for the QOT problem has sparse support for small regularization parameter ϵ>0.\epsilon>0. In this article we provide the first quantitative description of this phenomenon in general dimension: we derive bounds on the size and on the location of the support of πϵ\pi_\epsilon compared to the Monge coupling. Our analysis is based on pointwise bounds on the density of πϵ\pi_\epsilon together with Minty's trick, which provides a quadratic detachment from the optimal transport duality gap. In the self-transport setting μ=ν\mu=\nu we obtain optimal rates of order ϵ12+d.\epsilon^{\frac{1}{2+d}}.

Keywords

Cite

@article{arxiv.2410.03425,
  title  = {Sparsity of Quadratically Regularized Optimal Transport: Bounds on concentration and bias},
  author = {Johannes Wiesel and Xingyu Xu},
  journal= {arXiv preprint arXiv:2410.03425},
  year   = {2024}
}