Skip to content
DraconiticResearch
Search Sources API
203,234 studies31 sourcesUpdated 32 min ago
PreprintCC BY 4.0arXiv

Accelerating Sinkhorn for Entropy-Regularized Optimal Transport

  • Long Chen
  • Zeyi Xu
Open at sourcePDFDOICite

Abstract

Entropy-regularized optimal transport is widely used to compare and align data distributions, but Sinkhorn can converge slowly at small regularization. We introduce Acc-Sinkhorn, a two-state Hessian-driven Nesterov iteration built on the dual mirror geometry of the row-reduced Sinkhorn map. The method combines momentum acceleration with an inexpensive descent safeguard, sta- ble log-domain evaluation, and a prescribed homotopy schedule requiring neither line search nor curvature estimation. We prove global linear convergence of the safeguarded method, rule out con- secutive restarts, and establish local accelerated convergence for fixed parameters without restarts through spectral analysis. Numerical comparisons with six methods on synthetic problems, color transfer, and bilingual word alignment show runtime speedups of up to about 26 over Sinkhorn, with additional rendering-accuracy comparisons for color transfer.

Related

  1. Exponential convergence of Sinkhorn algorithm for entropy martingale optimal transport
    Anna Kazeykina, Zhenjie Ren, Hecheng Wang
    arXivPreprintarXiv licenceFull textOptimisation & controlProbabilityarXiv
  2. Entropy-Regularized Optimal Transport for Time-Varying Multi-Agent Coverage Control
    Italo Napolitano, Mario di Bernardo
    arXivPreprintCC BY-NC-SA 4.0Full textOptimisation & controlarXiv
  3. Partial identification with entropy regularized optimal transport
    Bruno N. Costa, Florian F. Gunsilius
    arXivPreprintCC BY 4.0Full textStatisticsarXiv
  4. Beyond Empirical Support: Structured Outlier Generation via Sinkhorn Optimal Transport
    Haixiang Sun, Andrew L. Liu
    arXivPreprintCC BY 4.0Full textOptimisation & controlStatisticsarXiv
  5. Forward Tangent-Centered Sinkhorn DRO: Exact Nominal Recovery from Optimal Transport to KL
    Junjie Guo
    arXivPreprintCC BY 4.0Full textOptimisation & controlarXiv

Cite

BibTeXRIS
BibTeX
@misc{chen2026accelerating,
  title = {{Accelerating Sinkhorn for Entropy-Regularized Optimal Transport}},
  author = {Long Chen and Zeyi Xu},
  howpublished = {arXiv},
  year = {2026},
  month = nov,
  doi = {10.48550/arxiv.2605.30267},
  eprint = {2605.30267},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2605.30267},
  note = {CC BY 4.0},
}
RIS
TY  - UNPB
TI  - Accelerating Sinkhorn for Entropy-Regularized Optimal Transport
AU  - Long Chen
AU  - Zeyi Xu
PY  - 2026
DA  - 2026/11/02
PB  - arXiv
DO  - 10.48550/arxiv.2605.30267
UR  - https://arxiv.org/abs/2605.30267
AB  - Entropy-regularized optimal transport is widely used to compare and align data distributions, but Sinkhorn can converge slowly at small regularization. We introduce Acc-Sinkhorn, a two-state Hessian-driven Nesterov iteration built on the dual mirror geometry of the row-reduced Sinkhorn map. The method combines momentum acceleration with an inexpensive descent safeguard, sta- ble log-domain evaluation, and a prescribed homotopy schedule requiring neither line search nor curvature estimation. We prove global linear convergence of the safeguarded method, rule out con- secutive restarts, and establish local accelerated convergence for fixed parameters without restarts through spectral analysis. Numerical comparisons with six methods on synthetic problems, color transfer, and bilingual word alignment show runtime speedups of up to about 26 over Sinkhorn, with additional rendering-accuracy comparisons for color transfer.
ER  -