New dual formulation clarifies sample complexity for unbalanced entropic OT

This paper introduces a new dual formulation for unbalanced entropic optimal transport (OT), focusing on its sample complexity at the optimal coupling level. The research demonstrates that entropic regularization is crucial for machine learning applications, as it mitigates the curse of dimensionality, reduces sample requirements for stable transport estimation, and ensures compatibility with efficient Sinkhorn-type algorithms. The findings provide theoretical backing for the practical necessity of these methods in handling noisy empirical data and complex probability distributions. AI

IMPACT Provides theoretical justification for using entropic regularization in machine learning, potentially improving model stability and efficiency.

RANK_REASON Academic paper detailing a new theoretical formulation and its implications for machine learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• Connected Papers
• CORE Recommender
• DagsHub
• entropic unbalanced OT
• Gotit.pub
• Hugging Face
• Litmaps
• machine learning
• optimal transport
• ScienceCast
• scite Smart Citations
• Sinkhorn-type algorithms

• paper
• other

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