Convergence of empirical subgradients for optimal transport-based objectives

Tam Le

Optimal transport is widely used to learn distributions, enforce distributional constraints, and model uncertainty. In applications, transport losses are often computed from sample…