On the Efficiency of Entropic Regularized Algorithms for Optimal Transport

TL;DR

This paper advances the understanding of entropic regularized algorithms for optimal transport by improving complexity bounds, proposing new algorithms, and demonstrating their practical efficiency through experiments.

Contribution

It provides new complexity bounds for Greenkhorn and introduces APDAMD, a generalized accelerated primal-dual algorithm, along with a deterministic accelerated Sinkhorn variant.

Findings

Greenkhorn outperforms Sinkhorn in practice due to improved complexity bounds.

APDAMD achieves a complexity bound of rac{n^2\u221a{\u03b4}}{\u03b5} with regularity b4.

Accelerated Sinkhorn outperforms standard Sinkhorn and Greenkhorn in terms of b5 and n.

Abstract

We present several new complexity results for the entropic regularized algorithms that approximately solve the optimal transport (OT) problem between two discrete probability measures with at most $n$ atoms. First, we improve the complexity bound of a greedy variant of Sinkhorn, known as \textit{Greenkhorn}, from $\widetilde{O}(n^2\varepsilon^{-3})$ to $\widetilde{O}(n^2\varepsilon^{-2})$. Notably, our result can match the best known complexity bound of Sinkhorn and help clarify why Greenkhorn significantly outperforms Sinkhorn in practice in terms of row/column updates as observed by~\citet{Altschuler-2017-Near}. Second, we propose a new algorithm, which we refer to as \textit{APDAMD} and which generalizes an adaptive primal-dual accelerated gradient descent (APDAGD) algorithm~\citep{Dvurechensky-2018-Computational} with a prespecified mirror mapping $\phi$. We prove that APDAMD achieves the complexity bound of $\widetilde{O}(n^2\sqrt{\delta}\varepsilon^{-1})$ in which $\delta>0$ stands for the regularity of $\phi$. In addition, we show by a counterexample that the complexity bound of $\widetilde{O}(\min\{n^{9/4}\varepsilon^{-1}, n^2\varepsilon^{-2}\})$ proved for APDAGD before is invalid and give a refined complexity bound of $\widetilde{O}(n^{5/2}\varepsilon^{-1})$. Further, we develop a \textit{deterministic} accelerated variant of Sinkhorn via appeal to estimated sequence and prove the complexity bound of $\widetilde{O}(n^{7/3}\varepsilon^{-4/3})$. As such, we see that accelerated variant of Sinkhorn outperforms Sinkhorn and Greenkhorn in terms of $1/\varepsilon$ and APDAGD and accelerated alternating minimization (AAM)~\citep{Guminov-2021-Combination} in terms of $n$. Finally, we conduct the experiments on synthetic and real data and the numerical results show the efficiency of Greenkhorn, APDAMD and accelerated Sinkhorn in practice.