English

Optimal transport problems regularized by generic convex functions: A geometric and algorithmic approach

Optimization and Control 2020-11-30 v1

Abstract

In order to circumvent the difficulties in solving numerically the discrete optimal transport problem, in which one minimizes the linear target function PC,P:=i,jCijPijP\mapsto\langle C,P\rangle:=\sum_{i,j}C_{ij}P_{ij}, Cuturi introduced a variant of the problem in which the target function is altered by a convex one Φ(P)=C,PλH(P)\Phi(P)=\langle C,P\rangle-\lambda\mathcal{H}(P), where H\mathcal{H} is the Shannon entropy and λ\lambda is a positive constant. We herein generalize their formulation to a target function of the form Φ(P)=C,P+λf(P)\Phi(P)=\langle C,P\rangle+\lambda f(P), where ff is a generic strictly convex smooth function. We also propose an iterative method for finding a numerical solution, and clarify that the proposed method is particularly efficient when f(P)=12P2f(P)=\frac{1}{2}\|P\|^2.

Keywords

Cite

@article{arxiv.2011.13683,
  title  = {Optimal transport problems regularized by generic convex functions: A geometric and algorithmic approach},
  author = {Daiji Tsutsui},
  journal= {arXiv preprint arXiv:2011.13683},
  year   = {2020}
}
R2 v1 2026-06-23T20:32:59.621Z