计算机科学

计算机科学 ›› 2026, Vol. 53 ›› Issue (3): 173-180.doi: 10.11896/jsjkx.250200111

• 数据库 & 大数据 & 数据科学 • 上一篇    下一篇

基于机械遗忘的部分域自适应

吴嘉豪, 彭力, 杨杰龙   

  1. 江南大学物联网工程学院物联网技术应用教育部工程研究中心 江苏 无锡 214122
  • 收稿日期:2025-02-26 修回日期:2025-04-29 发布日期:2026-03-12
  • 通讯作者: 彭力(penglimail2002@163.com)
  • 作者简介:(6231913054@stu.jiangnan.edu.cn)
  • 基金资助:
    国家自然科学基金(61873112)

Partial Domain Adaptation Based on Machine Unlearning

WU Jiahao, PENG Li, YANG Jielong   

  1. Ministry of Education Engineering Research Center for the Application of Internet of Things Technology, School of Internet of Things Enginee-ring, Jiangnan University, Wuxi, Jiangsu 214122, China
  • Received:2025-02-26 Revised:2025-04-29 Online:2026-03-12
  • About author:WU Jiahao,born in 2002,postgraduate.His main research interests include transfer learning and computer vision.
    PENG Li,born in 1967,Ph.D,professor,Ph.D supervisor.His main research interests include computer vision,deep learning and visual Internet of Things.
  • Supported by:
    National Natural Science Foundation of China(61873112).

PDF (PC)

103

摘要： 域适应将知识从标签丰富的源领域转移到标签稀缺的目标领域,在减少目标领域数据标注需求的情况下,实现模型性能在目标领域的提升。作为一种更现实的扩展,部分域自适应放宽了源领域和目标领域完全共享标签空间的假设,并处理目标标签空间是源标签空间子集的情况。所提出的机械遗忘方法,通过遗忘异常权重类别来帮助解决具有挑战性的部分域自适应问题。具体而言,该方法首先采用传统部分域适应方法作为初始化模型,同时通过类别权重机制识别出异常权重类别;然后根据异常权重类别筛选源域数据集并生成噪声样本数据集,进而对模型进行遗忘操作,解决源域和目标域标签空间不匹配的问题;最后利用伪标签技术,让模型进一步对齐目标域的特征分布,从而促进正迁移。在Office-31和Office-Home这两个公开的基准数据集上进行的大量实验表明,所提出的机械遗忘方法在与最新的部分域自适应方法的性能相近的同时,显著超过了传统的部分域适应方法。

关键词: 迁移学习, 部分域自适应, 机械遗忘, 伪标签, 负迁移

Abstract: Domain adaptation aims to transfer knowledge from a label-rich source domain to a label-scarce target domain,thereby improving model performance on the target domain while reducing the need for target domain data annotation.As a more realistic extension,partial domain adaptation(PDA) relaxes the assumption of complete label space sharing between the source and target domains,and handles the case where the target label space is a subset of the source label space.The proposed machine unlearning method helps to address the challenging problem of partial domain adaptation by forgetting outlier-weighted categories.Specifically,the method firstly uses a traditional PDA method as the initialization model,while the category weight mechanism identifies the outlier-weighted categories.Then,it screens the source domain dataset based on the outlier-weighted categories and generates a noisy sample dataset,and then performs unlearning on the model to solve the label space mismatch problem between the source and target domains.Finally,it leverages pseudo-labeling techniques to further align the feature distribution of the target domain,thereby promoting positive transfer.Extensive experiments on the publicly available Office-31 and Office-Home benchmark datasets show that the proposed machine unlearning method not only performs on par with the latest PDA methods,but also significantly outperforms traditional PDA methods.

Key words: Transfer learning, Partial domain adaptation, Machine unlearning, Pseudo label, Negative transfer

中图分类号: 

  • TP391
[1]FARAHANI A,VOGHOEI S,RASHEED K,et al.A brief review of domain adaptation[C]//Advances in Data Science and Information Engineering:Proceedings from ICDATA 2020 and IKE 2020.2021:877-894.
[2]YOU K,LONG M,CAO Z,et al.Universal domain adaptation[C]//Proceedings of the IEEE/CVFConference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2019:2720-2729.
[3]TIAN Y,ZHU S.Partial domain adaptation on semantic seg-mentation[J].IEEE Transactions on Circuits and Systems for Video Technology,2021,32(6):3798-3809.
[4]PENG X,BAI Q,XIA X,et al.Moment matching for multi-source domain adaptation[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.IEEE,2019:1406-1415.
[5]CAO Z,MA L,LONG M,et al.Partial adversarial domain adaptation[C]//Proceedings of the European Conference on Computer Vision.Berlin:Springer,2018:135-150.
[6]SANKARANARAYANAN S,BALAJI Y,CASTILLO C D,et al.Generate to adapt:Aligning domains using generative adversarial networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2018:8503-8512.
[7]CHEN J,WU X,DUAN L,et al.Domain adversarial reinforcement learning for partial domain adaptation[J].IEEE Transactions on Neural Networks and Learning Systems,2020,33(2):539-553.
[8]CHEN Z,CHEN C,CHENG Z,et al.Selective transfer with reinforced transfer network for partial domain adaptation[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2020:12706-12714.
[9]WU K,WU M,YANG J,et al.Deep Reinforcement LearningBoosted Partial Domain Adaptation[C]//IJCAI.San Francisco,CA:Morgan Kaufmann Publishers,2021:3192-3199.
[10]CAO Z,LONG M,WANG J,et al.Partial transfer learning with selective adversarial networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Pisca-taway,NJ:IEEE,2018:2724-2732.
[11]ZHANG J,DING Z,LI W,et al.Importance weighted adversarial nets for partial domain adaptation[C]//Proceedings of the IEEEConference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2018:8156-8164.
[12]LIANG J,WANG Y,HU D,et al.A balanced and uncertainty-aware approach for partial domain adaptation[C]//EuropeanConference on Computer Vision.Cham:Springer,2020:123-140.
[13]LI S,GONG K,XIE B,et al.Critical classes and samples disco-vering for partial domain adaptation[J].IEEE Transactions on Cybernetics,2022,53(9):5641-5654.
[14]GANIN Y,USTINOVA E,AJAKAN H,et al.Domain-adversarial training of neural networks[J].Journal of Machine Lear-ning Research,2016,17(59):1-35.
[15]YANG J,QIAN H,XU Y,et al.Can We Evaluate Domain Adaptation Models Without Target-Domain Labels?[J].arXiv:2305.18712,2023.
[16]ZHAO H,DES COMBES R T,ZHANG K,et al.On learning invariant representations for domain adaptation[C]//Internatio-nal Conference on Machine Learning.Long Beach,CA:PMLR,2019:7523-7532.
[17]TZENG E,HOFFMAN J,SAENKO K,et al.Adversarial dis-criminative domain adaptation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2017:7167-7176.
[18]BERTHELOT D,CARLINI N,GOODFELLOW I,et al.Mix-match:A holistic approach to semi-supervised learning[J].Advances in Neural Information Processing Systems,2019,32:5049-5059.
[19]KIM Y,CHO D,HAN K,et al.Domain adaptation withoutsource data[J].IEEE Transactions on Artificial Intelligence,2021,2(6):508-518.
[20]KIM Y,CHO D,HONG S.Towards privacy-preserving domain adaptation[J].IEEE Signal Processing Letters,2020,27:1675-1679.
[21]XU X,HE H,ZHANG H,et al.Unsupervised domain adaptation via importance sampling[J].IEEE Transactions on Circuits and Systems for Video Technology,2019,30(12):4688-4699.
[22]ZHANG C,ZHAO Q.Attention guided for partial domain adaptation[J].Information Sciences,2021,547:860-869.
[23]LI S,LIU C H,LIN Q,et al.Deep residual correction network for partial domain adaptation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2020,43(7):2329-2344.
[24]CAO Z,YOU K,LONG M,et al.Learning to transfer examples for partial domain adaptation[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2019:2985-2994.
[25]GU X,YU X,SUN J,et al.Adversarial reweighting for partial domain adaptation[J].Advances in Neural Information Proces-sing Systems,2021,34:14860-14872.
[26]WU K,WU M,CHEN Z,et al.Reinforced adaptation network for partial domain adaptation[J].IEEE Transactions on Circuits and Systems for Video Technology,2022,33(5):2370-2380.
[27]LAN M,MENG M,YU J,et al.Learning to Discover Know-ledge:A Weakly-Supervised Partial Domain Adaptation Approach[J].IEEE Transactions on Image Processing,2024,33:4090-4103.
[28]LI G,HSU H,MARCULESCU R.Machine unlearning forimage-to-image generative models[J].arXiv:2402.00351,2024.
[29]NGUYEN T T,HUYNH T T,REN Z,et al.A survey of machine unlearning[J].arXiv:2209.02299,2022.
[30]WANG Z,YANG E,SHEN L,et al.A comprehensive survey of forgetting in deep learning beyond continual learning[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2024,47:1464-1483.
[31]DANG Q V.Right to be forgotten in the age of machine learning[C]//Advances in Digital Science(ICADS 2021).Cham:Sprin-ger,2021:403-411.
[32]BOURTOULE L,CHANDRASEKARAN V,CHOQUETTE-CHOO C A,et al.Machine unlearning[C]//2021 IEEE Symposium on Security and Privacy(SP).IEEE,2021:141-159.
[33]ILHARCO G,RIBEIRO M T,WORTSMAN M,et al.Editing models with task arithmetic[J].arXiv:2212.04089,2022.
[34]TARUN A K,CHUNDAWAT V S,MANDAL M,et al.Fast yet effective machine unlearning[J].IEEE Transactions on Neural Networks and Learning Systems,2023,35:13046-13055.
[35]SAENKO K,KULIS B,FRITZ M,et al.Adapting visual category models to new domains[C]//Computer Vision－ECCV 2010:11th EuropeanConference on Computer Vision.Berlin:Springer,2010:213-226.
[36]VENKATESWARA H,EUSEBIO J,CHAKRABORTY S,et al.Deep hashing network for unsupervised domain adaptation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2017:5018-5027.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市两江新区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发