A Unified Domain Adaptation Framework with Distinctive Divergence Analysis
Published in Transactions on Machine Learning Research, 2022
In this paper, we propose a unified framework that offers a shared understanding of domain adaptation problems, including both classification and regression, from the same perspective. We provide rigorous theoretical analysis as well as solid experimental evidence to support our framework, which can easily incorporate a large family of $f$-divergences. Through experiments on classical domain adaptation benchmarks, we demonstrate the superiority of our proposed method.