Adversarial Reweighting with $\alpha$-Power Maximization for Domain Adaptation

TL;DR

This paper introduces ARPM, a novel adversarial reweighting approach with $oldsymbol{ extalpha}$-power maximization for partial domain adaptation, effectively reducing negative transfer and improving classification on target domains.

Contribution

The paper proposes a new adversarial reweighting model with $oldsymbol{ extalpha}$-power maximization for partial domain adaptation, extending its application to open-set, universal, and test-time adaptation.

Findings

ARPM outperforms recent PDA methods on five benchmarks.

The $oldsymbol{ extalpha}$-power maximization reduces prediction uncertainty.

Theoretical analysis shows the method approximately minimizes target error bound.

Abstract

The practical Domain Adaptation (DA) tasks, e.g., Partial DA (PDA), open-set DA, universal DA, and test-time adaptation, have gained increasing attention in the machine learning community. In this paper, we propose a novel approach, dubbed Adversarial Reweighting with $\alpha$-Power Maximization (ARPM), for PDA where the source domain contains private classes absent in target domain. In ARPM, we propose a novel adversarial reweighting model that adversarially learns to reweight source domain data to identify source-private class samples by assigning smaller weights to them, for mitigating potential negative transfer. Based on the adversarial reweighting, we train the transferable recognition model on the reweighted source distribution to be able to classify common class data. To reduce the prediction uncertainty of the recognition model on the target domain for PDA, we present an $\alpha$-power maximization mechanism in ARPM, which enriches the family of losses for reducing the prediction uncertainty for PDA. Extensive experimental results on five PDA benchmarks, i.e., Office-31, Office-Home, VisDA-2017, ImageNet-Caltech, and DomainNet, show that our method is superior to recent PDA methods. Ablation studies also confirm the effectiveness of components in our approach. To theoretically analyze our method, we deduce an upper bound of target domain expected error for PDA, which is approximately minimized in our approach. We further extend ARPM to open-set DA, universal DA, and test time adaptation, and verify the usefulness through experiments.