Domain Adaptation with Adversarial Training on Penultimate Activations
Open Access
- 26 June 2023
- journal article
- research article
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 37 (8) , 9935-9943
Abstract
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.Keywords
This publication has 3 references indexed in Scilit:
- A Survey of Unsupervised Deep Domain AdaptationACM Transactions on Intelligent Systems and Technology, 2020
- Deep visual domain adaptation: A surveyNeurocomputing, 2018
- A Survey on Transfer LearningIEEE Transactions on Knowledge and Data Engineering, 2009