Discriminative Adversarial Domain Adaptation

TL;DR

This paper introduces Discriminative Adversarial Domain Adaptation (DADA), a novel method that improves alignment of joint feature-category distributions across domains by using an integrated classifier and a unique adversarial objective, advancing unsupervised domain adaptation.

Contribution

The paper proposes DADA, a new adversarial learning approach with an integrated classifier and a minimax game, addressing mode collapse and extending to partial and open set domain adaptation.

Findings

Achieves state-of-the-art results on benchmark datasets for all three adaptation settings.

Effectively aligns joint feature-category distributions across domains.

Demonstrates robustness in partial and open set domain adaptation scenarios.

Abstract

Given labeled instances on a source domain and unlabeled ones on a target domain, unsupervised domain adaptation aims to learn a task classifier that can well classify target instances. Recent advances rely on domain-adversarial training of deep networks to learn domain-invariant features. However, due to an issue of mode collapse induced by the separate design of task and domain classifiers, these methods are limited in aligning the joint distributions of feature and category across domains. To overcome it, we propose a novel adversarial learning method termed Discriminative Adversarial Domain Adaptation (DADA). Based on an integrated category and domain classifier, DADA has a novel adversarial objective that encourages a mutually inhibitory relation between category and domain predictions for any input instance. We show that under practical conditions, it defines a minimax game that can promote the joint distribution alignment. Except for the traditional closed set domain adaptation, we also extend DADA for extremely challenging problem settings of partial and open set domain adaptation. Experiments show the efficacy of our proposed methods and we achieve the new state of the art for all the three settings on benchmark datasets.