Generalization Bounds with Minimal Dependency on Hypothesis Class via Distributionally Robust Optimization

In contrast to the hypothesis class complexity in ERM, our DRO bounds depend on the ambiguity set geometry and its compatibility with the true loss function.