Towards Backwards-Compatible Data with Confounded Domain Adaptation

TL;DR

This paper introduces a novel framework for achieving backwards-compatible data across domains with confounded shifts, enabling versatile downstream use without retraining models.

Contribution

It proposes a new approach to handle confounded domain shifts, extending generalized label shift to improve data compatibility for multiple tasks.

Findings

Effective on synthetic datasets

Demonstrates improved data compatibility

Applicable to real-world datasets

Abstract

Most current domain adaptation methods address either covariate shift or label shift, but are not applicable where they occur simultaneously and are confounded with each other. Domain adaptation approaches which do account for such confounding are designed to adapt covariates to optimally predict a particular label whose shift is confounded with covariate shift. In this paper, we instead seek to achieve general-purpose data backwards compatibility. This would allow the adapted covariates to be used for a variety of downstream problems, including on pre-existing prediction models and on data analytics tasks. To do this we consider a modification of generalized label shift (GLS), which we call confounded shift. We present a novel framework for this problem, based on minimizing the expected divergence between the source and target conditional distributions, conditioning on possible confounders. Within this framework, we provide concrete implementations using the Gaussian reverse Kullback-Leibler divergence and the maximum mean discrepancy. Finally, we demonstrate our approach on synthetic and real datasets.