Environment Inference for Invariant Learning

               Elliot Creager 1 2 Jorn-Henrik Jacobsen 1 2 Richard Zemel 1 2

             Abstract
    Learning models that gracefully handle distribu-
    tion shifts is central to research on domain gen-
    eralization, robust optimization, and fairness. A
    promising formulation is domain-invariant learn-
    ing, which identifies the key issue of learning
    which features are domain-specific versus domain-
    invariant. An important assum ...