Weighted and Class-Specific Maximum Mean Discrepancy for Unsupervised Domain Adaptation

被引:57
|
作者
Yan, Hongliang [1 ]
Li, Zhetao [2 ,3 ]
Wang, Qilong [4 ]
Li, Peihua [5 ]
Xu, Yong [6 ]
Zuo, Wangmeng [1 ]
机构
[1] Harbin Inst Technol, Sch Comp Sci & Technol, Harbin 150001, Peoples R China
[2] Xiangtan Univ, Key Lab Hunan Prov Internet Things & Informat Sec, Xiangtan 411105, Hunan, Peoples R China
[3] Xiangtan Univ, Coll Informat Engn, Xiangtan 411105, Hunan, Peoples R China
[4] Tianjin Univ, Coll Intelligence & Comp, Tianjin 300350, Peoples R China
[5] Dalian Univ Technol, Sch Informat & Commun Engn, Dalian 116024, Peoples R China
[6] Harbin Inst Technol, Biocomp Res Ctr, Shenzhen 518055, Peoples R China
基金
国家重点研发计划;
关键词
Measurement; Adaptation models; Airplanes; Gallium nitride; Task analysis; Generative adversarial networks; Degradation; Image recognition; unsupervised domain adaption; convolutional neural network; expectation-maximization algorithms; KERNEL; TEXT;
D O I
10.1109/TMM.2019.2953375
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Although maximum mean discrepancy (MMD) has achieved great success in unsupervised domain adaptation (UDA), most of existing UDA methods ignore the issue of class weight bias across domains, which is ubiquitous and evidently gives rise to the degradation of UDA performance. In this work, we propose two improved MMD metrics, i.e., weighted MMD (WMMD) and class-specific MMD (CMMD), to alleviate the adverse effect caused by the changes of class prior distributions between source and target domains. In WMMD, class-specific auxiliary weights are deployed to reweigh the source samples. In CMMD, we calculate the MMD for each class of source and target samples. Since the class labels of target samples are unknown for UDA problem, we present a classification expectation-maximization algorithm to estimate the pseudo-labels of target samples on the fly and update the model parameters using estimated labels. The proposed methods can be flexibly incorporated into deep convolutional neural networks to form WMMD and CMMD based domain adaptation networks, which we called WDAN and CDAN, respectively. By combining WMMD with CMMD, we present a CWMMD based domain adaptation network (CWDAN) to further improve classification performance. Experiments show that, both WMMD and CMMD benefit the classification accuracy, and our CWDAN can achieve compelling UDA performance in comparison with MMD and the state-of-the-art UDA methods.
引用
收藏
页码:2420 / 2433
页数:14
相关论文
共 50 条
  • [1] Mind the Class Weight Bias: Weighted Maximum Mean Discrepancy for Unsupervised Domain Adaptation
    Yan, Hongliang
    Ding, Yukang
    Li, Peihua
    Wang, Qilong
    Xu, Yong
    Zuo, Wangmeng
    [J]. 30TH IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2017), 2017, : 945 - 954
  • [2] Maximum Mean and Covariance Discrepancy for Unsupervised Domain Adaptation
    Wenju Zhang
    Xiang Zhang
    Long Lan
    Zhigang Luo
    [J]. Neural Processing Letters, 2020, 51 : 347 - 366
  • [3] Maximum Mean and Covariance Discrepancy for Unsupervised Domain Adaptation
    Zhang, Wenju
    Zhang, Xiang
    Lan, Long
    Luo, Zhigang
    [J]. NEURAL PROCESSING LETTERS, 2020, 51 (01) : 347 - 366
  • [4] Extreme Learning Machine Based on Maximum Weighted Mean Discrepancy for Unsupervised Domain Adaptation
    Si, Yanna
    Pu, Jiexin
    Zang, Shaofei
    Sun, Lifan
    [J]. IEEE ACCESS, 2021, 9 : 2283 - 2293
  • [5] Domain adaptation learning based on maximum distribution weighted mean discrepancy
    Zang, Shao-Fei
    Cheng, Yu-Hu
    Wang, Xue-Song
    [J]. Kongzhi yu Juece/Control and Decision, 2016, 31 (11): : 2083 - 2089
  • [6] Unsupervised Domain Adaptation using Maximum Mean Covariance Discrepancy and Variational Autoencoder
    Barreto, Fabian
    Sarvaiya, Jignesh
    Patnaik, Suprava
    Yadav, Sushilkumar
    [J]. INTERNATIONAL JOURNAL OF ADVANCED COMPUTER SCIENCE AND APPLICATIONS, 2022, 13 (06) : 883 - 891
  • [7] Unsupervised domain adaptation using maximum mean discrepancy optimization for lithology identification
    Chang, Ji
    Li, Jing
    Kang, Yu
    Lv, Wenjun
    Xu, Ting
    Li, Zerui
    Zheng, Wei Xing
    Han, Hongwei
    Liu, Haining
    [J]. GEOPHYSICS, 2021, 86 (02) : ID19 - ID30
  • [8] Class-specific regularized joint distribution alignment for unsupervised domain adaptation
    Luo, Tian-jian
    [J]. ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE, 2024, 131
  • [9] Maximum Classifier Discrepancy for Unsupervised Domain Adaptation
    Saito, Kuniaki
    Watanabe, Kohei
    Ushiku, Yoshitaka
    Harada, Tatsuya
    [J]. 2018 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2018, : 3723 - 3732
  • [10] Continual Unsupervised Domain Adaptation for Semantic Segmentation using a Class-Specific Transfer
    Marsden, Robert A.
    Wiewel, Felix
    Doebler, Mario
    Yang, Yang
    Yang, Bin
    [J]. 2022 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN), 2022,