math.OCAug 12, 2026

Tight Nonasymptotic Local Convergence of Sinkhorn-Knopp

Authors: Wenzhi GaoZhaonan QuYinyu YeMadeleine Udell

Abstract

We revisit the Sinkhorn-Knopp (SK) algorithm for the matrix scaling problem. Despite extensive literature on the global convergence of SK and its variants, its local linear convergence behavior remains less understood. We address this gap by providing the first nonasymptotic local analysis of SK that matches the rate obtained from existing asymptotic Jacobian-based arguments. We show that under certain connectivity conditions, SK is a polynomial-time algorithm for doubly stochastic matrix scaling. With the developed tools, we showcase the local suboptimality of SK and provide accelerated variants. Finally, for dense matrices, we improve the complexity of existing first-order matrix scaling algorithms from O(n7/3ε2/3)O(\tfrac{n^{7/3}}{\varepsilon^{2/3}}) to O(n9/4ε)O(\tfrac{n^{9/4}}{\sqrt{\varepsilon}}).

Explore similar work

CardsList
  1. Near-Optimal Nonconvex Matrix Completion

    Sep 15, 2026Jian-Feng Cai, Xiliang Lu, Juntao YouTensor CompletionConvex Optimization

  2. Well-Conditioned Oblivious Perturbations in Linear Space

    Apr 25, 2026Shabarish Chenakkod, Michał Dereziński, Xiaoyu Dong +1Spectral PreconditioningOptimal Sample Complexity