Class Overwhelms: Mutual Conditional Blended-Target Domain Adaptation

TL;DR

This paper introduces a novel mutual conditional alignment approach for blended target domain adaptation that effectively models categorical distributions without relying on domain labels, improving performance under label shift.

Contribution

It proposes a mutual reinforced mechanism using a categorical domain discriminator guided by uncertainty and low-level feature augmentation to enhance domain adaptation.

Findings

Outperforms state-of-the-art BTDA methods, especially under label distribution shift.

Effective in single target domain adaptation on DomainNet.

Does not require domain labels for alignment.

Abstract

Current methods of blended targets domain adaptation (BTDA) usually infer or consider domain label information but underemphasize hybrid categorical feature structures of targets, which yields limited performance, especially under the label distribution shift. We demonstrate that domain labels are not directly necessary for BTDA if categorical distributions of various domains are sufficiently aligned even facing the imbalance of domains and the label distribution shift of classes. However, we observe that the cluster assumption in BTDA does not comprehensively hold. The hybrid categorical feature space hinders the modeling of categorical distributions and the generation of reliable pseudo labels for categorical alignment. To address these, we propose a categorical domain discriminator guided by uncertainty to explicitly model and directly align categorical distributions $P(Z|Y)$. Simultaneously, we utilize the low-level features to augment the single source features with diverse target styles to rectify the biased classifier $P(Y|Z)$ among diverse targets. Such a mutual conditional alignment of $P(Z|Y)$ and $P(Y|Z)$ forms a mutual reinforced mechanism. Our approach outperforms the state-of-the-art in BTDA even compared with methods utilizing domain labels, especially under the label distribution shift, and in single target DA on DomainNet. Source codes are available at https://github.com/Pengchengpcx/Class-overwhelms-Mutual-Conditional-Blended-Target-Domain-Adaptation.