Towards Backwards-Compatible Data with Confounded Domain Adaptation
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 special case 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 propose using the reverse Kullback-Leibler divergence, demonstrating the use of parametric and nonparametric Gaussian estimators of the conditional distribution. We also propose using the Maximum Mean Discrepancy (MMD), introducing a dynamic strategy for choosing the kernel bandwidth, which is applicable even outside the confounded shift setting. Finally, we demonstrate our approach on synthetic and real datasets.
READ FULL TEXT