New analysis details Sinkhorn-Knopp algorithm's local convergence

Researchers have published a new analysis of the Sinkhorn-Knopp (SK) algorithm, focusing on its local convergence properties. The paper provides the first nonasymptotic local analysis of SK, matching existing asymptotic rates and demonstrating its polynomial-time solvability for doubly stochastic matrix scaling under specific connectivity conditions. The work also introduces accelerated variants and improves the complexity for dense matrices from $O( frac{n^{7/3}}{\varepsilon^{2/3}})$ to $O( frac{n^{9/4}}{\sqrt{\varepsilon}})$. AI

RANK_REASON Academic paper published on arXiv detailing a new analysis of a mathematical algorithm. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv stat.ML →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Influence Flower
• ScienceCast
• Sinkhorn-Knopp
• SK Hynix

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →