Imbalanced learning(IL),i.e.,learning unbiased models from class-imbalanced data, is a challenging problem.

To address this, we propose a new probabilistic framework for DG where the goal is to learn predictors that perform wellwith highprobability.

Since the VB algorithm does not depend on the unknown truth to achieve optimality, our results shed light on effective prior choices.

We thank the reviewers for their constructive suggestions. A summary of the added discussion is provided point-by-point below. Reviewer 1: How much more work is needed versus linear regression. The technical details are thus different (and more involved) here. We have included these derivations for completeness.