Moment Matching for Multi-Source Domain Adaptation

被引:757
|
作者
Peng, Xingchao [1 ]
Bai, Qinxun [2 ]
Xia, Xide [1 ]
Huang, Zijun [3 ]
Saenko, Kate [1 ]
Wang, Bo [4 ,5 ]
机构
[1] Boston Univ, Boston, MA 02215 USA
[2] Horizon Robot, Beijing, Peoples R China
[3] Columbia Univ, New York, NY 10027 USA
[4] Vector Inst, Toronto, ON, Canada
[5] Peter Munk Cardiac Ctr, Toronto, ON, Canada
关键词
D O I
10.1109/ICCV.2019.00149
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Conventional unsupervised domain adaptation (UDA) assumes that training data are sampled from a single domain. This neglects the more practical scenario where training data are collected from multiple sources, requiring multi-source domain adaptation. We make three major contributions towards addressing this problem. First, we collect and annotate by far the largest UDA dataset, called DomainNet, which contains six domains and about 0.6 million images distributed among 345 categories, addressing the gap in data availability for multi-source UDA research. Second, we propose a new deep learning approach, Moment Matching for Multi-Source Domain Adaptation (M(3)SDA), which aims to transfer knowledge learned from multiple labeled source domains to an unlabeled target domain by dynamically aligning moments of their feature distributions. Third, we provide new theoretical insights specifically for moment matching approaches in both single and multiple source domain adaptation. Extensive experiments are conducted to demonstrate the power of our new dataset in benchmarking state-of-the-art multi-source domain adaptation methods, as well as the advantage of our proposed model.
引用
收藏
页码:1406 / 1415
页数:10
相关论文
共 50 条
  • [31] Multi-source unsupervised domain adaptation for object detection
    Zhang, Dan
    Ye, Mao
    Liu, Yiguang
    Xiong, Lin
    Zhou, Lihua
    INFORMATION FUSION, 2022, 78 : 138 - 148
  • [32] STEM: An approach to Multi-source Domain Adaptation with Guarantees
    Nguyen, Van-Anh
    Nguyen, Tuan
    Le, Trung
    Tran, Quan Hung
    Phung, Dinh
    2021 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2021), 2021, : 9332 - 9343
  • [33] Riemannian representation learning for multi-source domain adaptation
    Chen, Sentao
    Zheng, Lin
    Wu, Hanrui
    PATTERN RECOGNITION, 2023, 137
  • [34] Multi-Source Domain Adaptation for Visual Sentiment Classification
    Lin, Chuang
    Zhao, Sicheng
    Meng, Lei
    Chua, Tat-Seng
    THIRTY-FOURTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, THE THIRTY-SECOND INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE AND THE TENTH AAAI SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE, 2020, 34 : 2661 - 2668
  • [35] Attention-Based Multi-Source Domain Adaptation
    Zuo, Yukun
    Yao, Hantao
    Xu, Changsheng
    IEEE TRANSACTIONS ON IMAGE PROCESSING, 2021, 30 : 3793 - 3803
  • [36] Improved multi-source domain adaptation by preservation of factors
    Schrom, Sebastian
    Hasler, Stephan
    Adamy, Juergen
    IMAGE AND VISION COMPUTING, 2021, 112
  • [37] Multi-Source Domain Adaptation and Fusion for Speaker Verification
    Zhu, Donghui
    Chen, Ning
    IEEE-ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING, 2022, 30 : 2103 - 2116
  • [38] Universal multi-Source domain adaptation for image classification
    Yin, Yueming
    Yang, Zhen
    Hu, Haifeng
    Wu, Xiaofu
    PATTERN RECOGNITION, 2022, 121
  • [39] Multi-source based approach for Visual Domain Adaptation
    Tiwari, Mrinalini
    Sanodiya, Rakesh Kumar
    Mathew, Jimson
    Saha, Sriparna
    2021 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN), 2021,
  • [40] Iterative Refinement for Multi-Source Visual Domain Adaptation
    Wu, Hanrui
    Yan, Yuguang
    Lin, Guosheng
    Yang, Min
    Ng, Michael K.
    Wu, Qingyao
    IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING, 2022, 34 (06) : 2810 - 2823