English

Representation Bayesian Risk Decompositions and Multi-Source Domain Adaptation

Machine Learning 2020-06-05 v2 Machine Learning

Abstract

We consider representation learning (hypothesis class H=FG\mathcal{H} = \mathcal{F}\circ\mathcal{G}) where training and test distributions can be different. Recent studies provide hints and failure examples for domain invariant representation learning, a common approach for this problem, but the explanations provided are somewhat different and do not provide a unified picture. In this paper, we provide new decompositions of risk which give finer-grained explanations and clarify potential generalization issues. For Single-Source Domain Adaptation, we give an exact decomposition (an equality) of the target risk, via a natural hybrid argument, as sum of three factors: (1) source risk, (2) representation conditional label divergence, and (3) representation covariate shift. We derive a similar decomposition for the Multi-Source case. These decompositions reveal factors (2) and (3) as the precise reasons for failure to generalize. For example, we demonstrate that domain adversarial neural networks (DANN) attempt to regularize for (3) but miss (2), while a recent technique Invariant Risk Minimization (IRM) attempts to account for (2) but does not consider (3). We also verify our observations experimentally.

Keywords

Cite

@article{arxiv.2004.10390,
  title  = {Representation Bayesian Risk Decompositions and Multi-Source Domain Adaptation},
  author = {Xi Wu and Yang Guo and Jiefeng Chen and Yingyu Liang and Somesh Jha and Prasad Chalasani},
  journal= {arXiv preprint arXiv:2004.10390},
  year   = {2020}
}

Comments

21 pages, 6 figures

R2 v1 2026-06-23T15:01:05.518Z