Double Robustness vs. Double Flexibility in Unsupervised Domain Adaptation

Abstract

Unsupervised domain adaptation seeks to transfer knowledge from a labeled source domain to an unlabeled target domain by addressing distributional discrepancies. In the literature, two key assumptions describe how these distributions differ: covariate shift and label shift. In this talk, I will introduce two related but distinct concepts under these assumptions: double robustness under covariate shift and double flexibility under label shift. Using techniques from semiparametric statistics, I will highlight their intrinsic similarities. Our findings shed light on the strengths and limitations of each paradigm, offering guidance for future research in robust and flexible domain adaptation strategies.

• Departments of Statistics and of Biostatistics & Medical Informatics [jiwei.zhao@wisc.edu]

Keywords: Double robustness – Double flexibility – Unsupervised domain adaptation