Domain Adaptation with Adversarial Training on Penultimate Activations
This addresses the challenge of improving domain adaptation performance for machine learning models, but it is incremental as it builds on existing adversarial training techniques.
The paper tackles the problem of enhancing model prediction confidence in Unsupervised Domain Adaptation by proposing adversarial training on penultimate activations, showing it is more efficient and better correlated with boosting confidence than prior methods, and achieves state-of-the-art scores on popular benchmarks.
Enhancing model prediction confidence on target data is an important objective in Unsupervised Domain Adaptation (UDA). In this paper, we explore adversarial training on penultimate activations, i.e., input features of the final linear classification layer. We show that this strategy is more efficient and better correlated with the objective of boosting prediction confidence than adversarial training on input images or intermediate features, as used in previous works. Furthermore, with activation normalization commonly used in domain adaptation to reduce domain gap, we derive two variants and systematically analyze the effects of normalization on our adversarial training. This is illustrated both in theory and through empirical analysis on real adaptation tasks. Extensive experiments are conducted on popular UDA benchmarks under both standard setting and source-data free setting. The results validate that our method achieves the best scores against previous arts. Code is available at https://github.com/tsun/APA.