Efficient Generalization with Distributionally Robust Learning

Consider a general formulation of the distributionally robust learning (DRL) problem of active interest.