Domain Confused Contrastive Learning for Unsupervised Domain Adaptation

TL;DR

This paper introduces Domain Confused Contrastive Learning (DCCL), a novel self-supervised method for unsupervised domain adaptation that improves domain invariance and discriminative features without target labels.

Contribution

DCCL is a new contrastive learning approach that uses domain puzzles and domain confused augmentations to enhance unsupervised domain adaptation.

Findings

DCCL significantly outperforms baseline methods in UDA tasks.

The proposed method effectively learns domain-invariant and discriminative representations.

Contrastive learning with domain puzzles improves adaptation performance.

Abstract

In this work, we study Unsupervised Domain Adaptation (UDA) in a challenging self-supervised approach. One of the difficulties is how to learn task discrimination in the absence of target labels. Unlike previous literature which directly aligns cross-domain distributions or leverages reverse gradient, we propose Domain Confused Contrastive Learning (DCCL) to bridge the source and the target domains via domain puzzles, and retain discriminative representations after adaptation. Technically, DCCL searches for a most domain-challenging direction and exquisitely crafts domain confused augmentations as positive pairs, then it contrastively encourages the model to pull representations towards the other domain, thus learning more stable and effective domain invariances. We also investigate whether contrastive learning necessarily helps with UDA when performing other data augmentations. Extensive experiments demonstrate that DCCL significantly outperforms baselines.