B0312
Title: ELSA: efficient label shift adaptation through the lens of semiparametric models
Authors: Jiwei Zhao - University of Wisconsin-Madison (United States) [presenting]
Abstract: The domain adaptation problem is studied with label shift. Under the label shift context, the marginal distribution of the label varies across the training and testing datasets, while the conditional distribution of features given the label is the same. Traditional label shift adaptation methods either suffer from large estimation errors or require cumbersome post-prediction calibrations. To address these issues, a moment-matching framework is first proposed for adapting the label shift based on the geometry of the influence function. Under such a framework, a novel method named efficient label shift adaptation (ELSA) is proposed, in which the adaptation weights can be estimated by solving linear systems. Theoretically, the ELSA estimator is root-n-consistent (n is the sample size of the source data) and asymptotically normal. Empirically, it is shown that ELSA can achieve state-of-the-art estimation performances without post-prediction calibrations, thus, gaining computational efficiency.
