Domain Adaptation with Adversarial Training on Penultimate Activations

被引:0
作者
Sun, Tao [1 ]
Lu, Cheng [2 ]
Ling, Haibin [1 ]
机构
[1] SUNY Stony Brook, Stony Brook, NY 11794 USA
[2] XPeng Motors, Guangzhou, Peoples R China
来源
THIRTY-SEVENTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOL 37 NO 8 | 2023年
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
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.
引用
收藏
页码:9935 / 9943
页数:9
相关论文
共 50 条
[41]   Adversarially Robust Source-free Domain Adaptation with Relaxed Adversarial Training [J].
Xiao, Yao ;
Wei, Pengxu ;
Liu, Cong ;
Lin, Liang .
2023 IEEE INTERNATIONAL CONFERENCE ON MULTIMEDIA AND EXPO, ICME, 2023, :2681-2686
[42]   Unsupervised Domain Adaptation for Dialogue Sequence Labeling Based on Hierarchical Adversarial Training [J].
Orihashi, Shota ;
Ihori, Mana ;
Tanaka, Tomohiro ;
Masumura, Ryo .
INTERSPEECH 2020, 2020, :1575-1579
[43]   Self-Training Based Adversarial Domain Adaptation for Radio Signal Recognition [J].
Liang, Zhi ;
Xie, Jian ;
Yang, Xin ;
Tao, Mingliang ;
Wang, Ling .
IEEE COMMUNICATIONS LETTERS, 2022, 26 (11) :2646-2650
[44]   Visualizing and Analyzing the Topology of Neuron Activations in Deep Adversarial Training [J].
Zhou, Youjia ;
Zhou, Yi ;
Ding, Jie ;
Wang, Bei .
TOPOLOGICAL, ALGEBRAIC AND GEOMETRIC LEARNING WORKSHOPS 2023, VOL 221, 2023, 221
[45]   Unsupervised Domain Adaptation for Dysarthric Speech Detection via Domain Adversarial Training and Mutual Information Minimization [J].
Wang, Disong ;
Deng, Liqun ;
Yeung, Yu Ting ;
Chen, Xiao ;
Liu, Xunying ;
Meng, Helen .
INTERSPEECH 2021, 2021, :2956-2960
[46]   Unsupervised domain adaptation with adversarial distribution adaptation network [J].
Zhou, Qiang ;
Zhou, Wen'an ;
Wang, Shirui ;
Xing, Ying .
NEURAL COMPUTING & APPLICATIONS, 2021, 33 (13) :7709-7721
[47]   Unsupervised domain adaptation with adversarial distribution adaptation network [J].
Qiang Zhou ;
Wen’an Zhou ;
Shirui Wang ;
Ying Xing .
Neural Computing and Applications, 2021, 33 :7709-7721
[48]   Adversarial Reweighting for Partial Domain Adaptation [J].
Gu, Xiang ;
Yu, Xi ;
Yang, Yan ;
Sun, Jian ;
Xu, Zongben .
ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 34 (NEURIPS 2021), 2021, 34
[49]   Adversarial Multiple Source Domain Adaptation [J].
Zhao, Han ;
Zhang, Shanghang ;
Wu, Guanhang ;
Costeira, Joao P. ;
Moura, Jose M. F. ;
Gordon, Geoffrey J. .
ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 31 (NIPS 2018), 2018, 31
[50]   Targeted adversarial discriminative domain adaptation [J].
Chen, Hua-Mei ;
Savakis, Andreas ;
Diehl, Ashley ;
Blasch, Erik ;
Wei, Sixiao ;
Chen, Genshe .
JOURNAL OF APPLIED REMOTE SENSING, 2021, 15 (03)