Connecting sufficient conditions for domain adaptation: source-guided uncertainty, relaxed divergences and discrepancy localization

TL;DR

This paper introduces a theoretical framework connecting relaxed divergence measures and localization techniques to improve domain adaptation, especially under label shift, by leveraging source domain structure.

Contribution

It provides a new risk bound for target domain adaptation and links divergence relaxation with localization, enhancing handling of label shift scenarios.

Findings

Theoretical justification for combining divergence relaxation and localization.

Improved handling of label shift in domain adaptation.

Connection between relaxed divergences and discrepancy localization.

Abstract

Recent advances in domain adaptation establish that requiring a low risk on the source domain and equal feature marginals degrade the adaptation's performance. At the same time, empirical evidence shows that incorporating an unsupervised target domain term that pushes decision boundaries away from the high-density regions, along with relaxed alignment, improves adaptation. In this paper, we theoretically justify such observations via a new bound on the target risk, and we connect two notions of relaxation for divergence, namely $\beta-$relaxed divergences and localization. This connection allows us to incorporate the source domain's categorical structure into the relaxation of the considered divergence, provably resulting in a better handling of the label shift case in particular.