Out-of-distribution generalization under random, dense distributional shifts

Many existing approaches for estimating parameters in settings with distributional shifts operate under an invariance assumption. For example, under covariate shift, it is assumed that p(y|x) remains invariant. We refer to such distribution shifts as sparse, since they may be substantial but affect only a part of the system. In contrast, in various real-world settings, shifts might be dense. More specifically, these shifts may arise through many small and random changes in the population and environment. First, we will discuss empirical evidence for such random dense distributional shifts and explain why commonly used models for distribution shifts — including adversarial approaches — may not be appropriate under these conditions. Then, we will develop tools to infer parameters and make predictions for partially observed, shifted distributions. Finally, we will apply the framework to several real-world datasets and discuss diagnostics to evaluate the fit of the distributional uncertainty model.

This is joint work with Yujin Jeong, Ying Jin, and Ivy Zhang.