计算机科学

计算机科学 ›› 2025, Vol. 52 ›› Issue (9): 212-219.doi: 10.11896/jsjkx.240700159

• 数据库&大数据&数据科学 • 上一篇    下一篇

基于伪标签不确定性估计的无源域自适应方法

黄超, 程春玲, 王有康   

  1. 南京邮电大学计算机学院、软件学院、网络空间安全学院 南京 210023
  • 收稿日期:2024-07-24 修回日期:2024-10-22 出版日期:2025-09-15 发布日期:2025-09-11
  • 通讯作者: 程春玲(chengcl@njupt.edu.cn)
  • 作者简介:(1222045633@njupt.edu.cn)
  • 基金资助:
    国家自然科学基金(61972201)

Source-free Domain Adaptation Method Based on Pseudo Label Uncertainty Estimation

HUANG Chao, CHENG Chunling, WANG Youkang   

  1. School of Computer Science,Nanjing University of Posts and Telecommunications,Nanjing 210023,China
  • Received:2024-07-24 Revised:2024-10-22 Online:2025-09-15 Published:2025-09-11
  • About author:HUANG Chao,born in 1999,postgra-duate.His main research interests include deep learning and domain adaptation.
    CHENG Chunling,born in 1972,professor.Her main research interests include data mining and data management.
  • Supported by:
    National Natural Science Foundation of China(61972201).

PDF (PC)

8

摘要： 无源域自适应大多采用基于伪标签的自监督学习方法来解决无源的问题,然而这些方法忽视了伪标签生成过程中,目标样本特征分布的聚类结构和分类决策边界处样本的不确定性对伪标签噪声的影响,降低了模型性能。为此,提出一种基于伪标签不确定性估计的无源域自适应方法。首先,对模型特征提取器参数进行多次扰动来模拟源知识被数据微调后的变化,并利用样本在不同扰动模型下的特征分布相似性来评估源知识的泛化不确定性;并提出通过极值信息熵来衡量目标域内部的隐含信息的不确定性,该信息熵根据预测概率中最大值与次最大值的数值差异采用不同的熵计算方法。其次,根据两种不确定性将目标样本分为可靠样本和不可靠样本,对可靠样本采用自监督学习,并以其预测概率结果为权重将样本特征更新至类原型中,同时,引入历史类原型以增强类原型的稳定性。对不可靠样本采用对比学习,使其靠近相似的类原型。在3个公开基准数据集Office-31,Office-Home和VisDA-C上与多个基线模型进行比较,提出的方法在分类准确度上得到提升,验证了其有效性。

关键词: 域自适应, 无源域数据, 不确定性估计, 信息熵, 对比学习

Abstract: Most of the source-free domain adaptive methods are self-supervised learning based on pseudo-labels to solve the source-free limitation.However,these methods overlook the impact of the clustering structure of target sample feature distributions and the uncertainty of samples near the decision boundary on pseudo-label noise during the generation process,which reduces the model’s performance.Therefore,this paper proposes a source-free domain adaptive method based on pseudo-label uncertainty estimation.Firstly,multiple perturbations are introduced to the model’s feature extractor parameters to simulate variations in source knowledge resulting from data fine-tuning.The similarity of target sample feature distributions under different perturbed models is utilized to assess the generalization uncertainty of the source knowledge.Moreover,an extreme value entropy is proposed to quantify the latent information uncertainty within the target domain,where distinct entropy calculation methods are applied based on the difference between the highest and second-highest prediction probabilities.Secondly,the target samples are divided into reliable samples and unreliable samples according to two kinds of uncertainty.For reliable samples,self-supervised learning is employed,using their prediction probabilities as weights to update sample features into class prototypes.Furthermore,historical class prototypes are incorporated to improve the stability of the class prototypes.Contrast learning is applied to unreliable samples to bring them closer to similar class prototypes.Compared with several baseline models,the classification accuracy is improved on three public benchmark datasets:Office-31,Office-Home and VisDA-C,and the effectiveness of the method is verified.

Key words: Domain adaptive, Source-free data, Uncertainty estimation, Information entropy, Contrastive Learning

中图分类号: 

  • TP391.1
[1]ZHANG L,FU J,WANG S,et al.Guide Subspace Learning for Unsupervised Domain Adaptation[J].IEEE Transactions on Neural Networks and Learning Systems,2019,31(9):3374-3388.
[2]TIAN J,ZHANG J,WEN L,et al.VDM-DA:Virtual DomainModeling for Source Data-Free Domain Adaptation[J].IEEE Transactions on Circuits and Systems for Video Technology,2021,32(6):3749-3760.
[3]LI R,JIAO Q,CAO W,et al.Model Adaptation:Unsupervised Domain Adaptation without Source Data[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE Computer Society,2020:9641-9650.
[4]LIANG J,HU D,FENG J.Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation[C]//Proceedings of International Conference on Machine Learning.New York:ACM,2020:6028-6039.
[5]YANG S,WANG Y,WEIJER J,et al.Exploiting the Intrinsic Neighborhood Structure for Source-Free Domain Adaptation[J].Advances in Neural Information Processing Systems,2021,34:29393-29405.
[6]LITRICO M,BUE A,MORERIO P.Guiding Pseudo-labels with Uncertainty Estimation for Source-free Unsupervised Domain Adaptation[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE Computer Society,2023:7640-7650.
[7]WU J,WANG G,GU R,et al.UPL-SFDA:Uncertainty-aware Pseudo Label Guided Source-Free DomainAdaptation for Medical Image Segmentation[J].IEEE Transactions on Medical Imaging,2023,42(12):3932-3943.
[8]CHEN X,YANG H,ZHANG H,et al.Uncertainty-AwarePseudo-Label Filtering for Source-Free Unsupervised Domain Adaptation[J].Neurocomputing,2024,575:127190.
[9]LI J,YU Z,DU Z,et al.A Comprehensive Survey on Source-free Domain Adaptation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2024,46(8):5743-5762.
[10]CUI C,ZHANG C,LIU Z,et al.Adversarial Source Generation for Source-Free Domain Adaptation[J].IEEE Transactions on Circuits and Systems for Video Technology,2024,34(6):4887-4898.
[11]HE J,WU L,TAO C,et al.Source-Free Domain Adaptationwith Unrestricted Source Hypothesis[J].Pattern Recognition,2024,149:110246.
[12]XIA H,ZHAO H,DING Z.Adaptive Adversarial Network for Source-free Domain Adaptation[C]//Proceedings of IEEE International Conference on Computer Vision.Piscataway,NJ:IEEE Computer Society,2021:9010-9019.
[13]TIAN Q,PENG S,MA T.Source-Free Unsupervised Domain Adaptation with Trusted Pseudo Samples[J].ACM Transactions on Intelligent Systems and Technology,2023,14(2):1-17.
[14]DING Y,SHENG L,LIANG J,et al.ProxyMix:Proxy-based Mixup Training with Label Refinery for Source-Free Domain Adaptation[J].Neural Networks,2023,167:92-103.
[15]KIM Y,CHO D,HAN K,et al.Domain Adaptation withoutSource Data[J].IEEE Transactions on Artificial Intelligence,2021,2(6):508-518.
[16]AHMED W,MORERIO P,MURINO V.Cleaning Noisy Labels by Negative Ensemble Learning for Source-Free Unsupervised Domain Adaptation[C]//Proceedings of IEEE Winter Confe-rence on Applications of Computer Vision.Piscataway,NJ:IEEE Computer Society,2022:1616-1625.
[17]KIM Y,YIM J,YUN J,et al.NLNL:Negative Learning forNoisy Labels[C]//Proceedings of IEEE International Confe-rence on Computer Vision.Piscataway,NJ:IEEE Computer So-ciety,2019:101-110.
[18]ZHANG C,BENGIO S,HARDT M,et al.Understanding Deep Learning (still) Requires Rethinking Generalization[J].Communications of the ACM,2021,64(3):107-115.
[19]MA X,HUANG H,WANG Y,et al.Normalized Loss Functions for Deep Learning with Noisy Labels[C]//Proceedings of International Conference on Machine Learning.New York:ACM,2020:6543-6553.
[20]WEI H,FENG L,CHEN X,et al.Combating Noisy Labels by Agreement:A Joint Training Method with Co-Regularization[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE Computer Society,2020:13726-13735.
[21]SAENKO K,KULIS B,FRITZ M,et al.Adapting Visual Category Models to New Domains[C]//Proceedings of European Conference on Computer Vision.Berlin:Springer,2010:213-226.
[22]VENKATESWARA H,EUSEBIO J,CHAKRABORTY S,et al.Deep Hashing Network for Unsupervised Domain Adaptation[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE Computer Society,2017:5018-5027.
[23]PENG X,USMAN B,KAUSHIK N,et al.VisDA:The Visual Domain Adaptation Challenge[J].arXiv:1710.06924,2017.
[24]GANIN Y,USTINOVA E,AJAKAN H,et al.Domain-Adversarial Training of Neural Networks[J].Journal of Machine Learning Research,2016,17(59):1-35.
[25]XU R,LI G,YANG J,et al.Larger norm more transferable:An adaptive feature norm approach for unsupervised domain adaptation[C]//Proceedings of IEEE International Conference on Computer Vision.Piscataway,NJ:IEEE Computer Society,2019:1426-1435.
[26]LIANG J,HU D,FENG J.Domain Adaptation with Auxiliary Target Domain-Oriented Classifier[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE Computer Society,2021:16632-16642.
[27]HE K,ZHANG X,REN S,et al.Deep Residual Learning for Image Recognition[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE Computer Society,2016:770-778.
[28]TANG S,SU W,YE M,et al.Source-Free Domain Adaptation with Frozen Multimodal Foundation Model[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE Computer Society,2024:23711-23720.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市渝北区洪湖西路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