Cycle Self-Training for Domain Adaptation

TL;DR

Cycle Self-Training (CST) is a novel algorithm for unsupervised domain adaptation that iteratively refines pseudo-labels across domains, improving target classification accuracy by enforcing cross-domain consistency and confidence regularization.

Contribution

This work introduces CST, a new self-training method with cycle consistency and Tsallis entropy regularization, providing theoretical analysis and outperforming existing methods in visual and sentiment domain adaptation.

Findings

CST recovers target ground truth in hard cases where other methods fail.

CST significantly outperforms state-of-the-art on visual recognition benchmarks.

CST improves sentiment analysis accuracy across domain shifts.

Abstract

Mainstream approaches for unsupervised domain adaptation (UDA) learn domain-invariant representations to narrow the domain shift. Recently, self-training has been gaining momentum in UDA, which exploits unlabeled target data by training with target pseudo-labels. However, as corroborated in this work, under distributional shift in UDA, the pseudo-labels can be unreliable in terms of their large discrepancy from target ground truth. Thereby, we propose Cycle Self-Training (CST), a principled self-training algorithm that explicitly enforces pseudo-labels to generalize across domains. CST cycles between a forward step and a reverse step until convergence. In the forward step, CST generates target pseudo-labels with a source-trained classifier. In the reverse step, CST trains a target classifier using target pseudo-labels, and then updates the shared representations to make the target classifier perform well on the source data. We introduce the Tsallis entropy as a confidence-friendly regularization to improve the quality of target pseudo-labels. We analyze CST theoretically under realistic assumptions, and provide hard cases where CST recovers target ground truth, while both invariant feature learning and vanilla self-training fail. Empirical results indicate that CST significantly improves over the state-of-the-arts on visual recognition and sentiment analysis benchmarks.