Distributional Robustness with IPMs and links to Regularization and GANs

Distributionally Robust Optimization (DRO), a particularly promising way of addressing this challenge, studies robustness via divergence-based uncertainty sets and considers robustness against shifts in distributions.