广东工业大学学报

• •    

置信样本选择与差异性特征增强的域适应

滕少华1, 吴泽锋1, 滕璐瑶2, 张巍1, 曾莹1   

  1. 1. 广东工业大学 计算机学院, 广东 广州 510006;
    2. 广州番禺职业技术学院 信息工程学院, 广东 广州 511483
  • 收稿日期:2023-11-21 出版日期:2024-09-27 发布日期:2024-09-27
  • 通信作者: 曾莹(1988–)，女，硕士研究生，主要研究方向为优化调度、协同计算、模式识别，E-mail：zengying@gdut.edu.cn
  • 作者简介:滕少华(1962–)，男，教授，博士生导师，CCF杰出会员，主要研究方向为数据挖掘、协同计算、模式识别，E-mail：shteng@gdut.edu.cn
  • 基金资助:
    国家自然科学基金资助项目(61972102)；广州市科技计划项目(2023A04J1729)

Confidence Sample Selection and Specific Feature Enhancement for Domain Adaptation

Teng Shao-hua1, Wu Ze-feng1, Teng Lu-yao2, Zhang Wei1, Zeng Ying1   

  1. 1. School of Computer Science and Technology, Guangdong University of Technology, Guangzhou 510006, China;
    2. School of Information Engineering, Guangzhou Panyu Polytechnic, Guangzhou 511483, China
  • Received:2023-11-21 Online:2024-09-27 Published:2024-09-27

HTML

PDF (PC)

20

摘要： 域适应学习因其有效减少域差异实现标签传播被广泛应用。目前，大多数域适应学习仅通过线性判别分析增强线性数据的判别性而忽略了现实世界存在非线性数据；同时，这些方法未考虑目标域的低置信样本对训练过程的负影响。因此，本文提出一种新颖的置信样本选择与差异性特征增强的域适应框架(Confidence Sample Selection and Specific Feature Enhancement, CSS-SFE)。首先，该框架通过最小最大原则选择高置信的目标样本来辅助训练，减少不正确伪标签影响；其次，权衡类散点矩阵和邻居散点矩阵的贡献来增强线性和非线性数据集的特征，提高数据集的判别性；接着，分别为源域和目标域的样本学习不同投影矩阵来保持各自的差异性特征，防止源域和目标域的样本区分性降低；再次，进一步应用边缘分布对齐和条件分布对齐减少域分布差异；最后，在多个基准数据集上进行的广泛实验证明该方法优于目前的方法。

关键词: 高置信样本选择, 差异性特征增强, 域适应, 标签传播

Abstract: For domain shifts, domain adaptation (DA) promotes label propagation by reducing distributional differences. In recent work, DA only enhances the discriminability of linear data by linear discriminant analysis, which ignores real-world nonlinear data. Meanwhile, these methods fail to consider the negative impact of the target low-confidence samples on the training process. Therefore, confidence sample selection and specific feature enhancement for domain adaptation (CSS-SFE) is proposed. Firstly, the framework selects target high-confidence samples through the min-max principle to reduce the impact of incorrect pseudo-label during training; Secondly, the class scatter matrix and neighbor scatter matrix are balanced to enhance the features of linear and nonlinear datasets so that the discriminability of the samples is improved; Thirdly, the framework maintains source and target specific features by learning different projection matrices, which prevents the discriminability of the samples decreasing; Fourthly, the marginal and conditional distribution alignment is further applied to reduce the domain distribution discrepancy; Finally, extensive experiments on several benchmark datasets demonstrate the superiority of CSS- SFE over state-of-the-art methods.

Key words: confidence sample selection, specific feature enhancement, domain adaptation, label propagation

中图分类号: 

  • TP181
[1] PAN S J, TSANG F W, KWOK J T, et al. Domain adaptation via transfer component analysis [J]. Journal of IEEE Transactions on Neural Networks and Learning Systems, 2011, 22: 199-210.
[2] LONG M S, WANG J M, DING G G, et al. Transfer feature learning with joint distribution adaptation[C]//2013 International Conference on Computer Vision. Sydney: IEEE, 2013: 2200-2207.
[3] GHIFARY M, BALDUZZI D, KLEIJN W B, et al. Scatter component analysis: a unified framework for domain adaptation and domain generalization [J]. Journal of IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39: 1414-1430.
[4] LI S, SONG S J, HUANG G, et al. Domain invariant and class discriminative feature learning for visual domain adaptation [J]. Journal of IEEE Transactions on Image Processing, 2018, 27: 4260-4273.
[5] ZHANG W, WU D R. Discriminative joint probability maximum mean discrepancy (DJP-MMD) for domain adaptation[C]//2020 International Joint Conference on Neural Networks. Glasgow, United Kingdom: IEEE, 2020, 1-8.
[6] TENG S H, ZHENG Z F, WU N Q, et al. Domain adaptation via incremental confidence samples into classification [J]. Journal of International Journal of Intelligent Systems, 2022, 37: 365-385.
[7] MENG M, LAN M C, YU J, et al. Dual-level adaptive and discriminative knowledge transfer for cross-domain recognition [J]. Journal of IEEE Transactions on Multimedia, 2023, 25: 2266-2279.
[8] ZHU F, GAO J B, YANG J, et al. Neighborhood linear discriminant analysis [J]. Journal of Pattern Recognition, 2022, 123: 108422.
[9] WANG J D, CHEN Y Q, HAO S J, et al. Balanced distribution adaptation for transfer learning[C]// 2017 International Conference on Data Mining. New Orleans: IEEE, 2017: 1129-1134.
[10] WANG J D, CHEN Y Q, YU H, et al. Easy transfer learning by exploiting intra-domain structures[C]//2019 International Conference on Multimedia and Expo. Shanghai: IEEE, 2019: 1210-1215.
[11] PENG Z H, ZHANG W, HAN N, et al. Active transfer learning [J]. Journal of IEEE Transactions on Circuits and Systems for Video Technology, 2020, 30: 1022-1036.
[12] TIAN L, TANG Y Q, HU Y Q, et al. Domain adaptation by class centroid matching and local manifold self-learning [J]. Journal of IEEE Transactions on Image Processing, 2020, 29: 9703-9718.
[13] XIAO N, ZHANG L, XU X, et al. Label disentangled analysis for unsupervised visual domain adaptation [J]. Journal of Knowledge-Based Systems, 2021, 229: 107309.
[14] LI Y, LI D S, LU Y W, et al. Progressive distribution alignment based on label correction for unsupervised domain adaptation[C]//2021 International Conference on Multimedia and Expo. Shenzhen: IEEE, 2021: 1-6.
[15] LU Y W, LI D S, WANG W J, et al. Discriminative invariant alignment for unsupervised domain adaptation [J]. Journal of IEEE Transactions on Multimedia, 2022, 24: 1871-1882.
[16] ZHAO J C, LI L S, DENG F, et al. Discriminant geometrical and statistical alignment with density peaks for domain adaptation [J]. Journal of IEEE Transactions on Cybernetics, 2022, 52: 1193-1206.
[17] WANG J D, FENG W J, CHEN Y Q, et al. Visual domain adaptation with manifold embedded distribution alignment[C]// 2018 Multimedia Conference on Multimedia Conference. Seoul: ACM, 2018: 402-410.
[18] ZHANG W, LI C, TENG S H. Joint discriminative distribution adaptation and manifold regularization for unsupervised domain adaptation[C]// 24th International Conference on Computer Supported Cooperative Work in Design. Dalian: IEEE, 2021: 226-231.
[19] YANG H W, HE H, ZHANG W Z, et al. Lie group manifold analysis: an unsupervised domain adaptation approach for image classification [J]. Journal of Applied Intelligence, 2022, 52: 4074-4088.
[20] SUN J, WANG Z H, WANG W, et al. Joint adaptive dual graph and feature selection for domain adaptation [J]. Journal of IEEE Transactions on Circuits and Systems for Video Technology, 2022, 32: 1453-1466.
[21] WANG Q, BRECKON. Unsupervised domain adaptation via structured prediction based selective pseudo-labeling[C]// 2020 Applications of Artificial Intelligence Conference. New York: AAAI, 2020: 6243-6250.
[22] ZHANG J, LI W Q, OGUNBONA P. Joint geometrical and statistical alignment for visual domain adaptation[C]//2017 Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 5150-5158.
[23] SARAY S N, TAHMORESNEZHAD J. Joint distinct subspace learning and unsupervised transfer classification for visual domain adaptation [J]. Journal of Signal Image Video Process, 2021, 15: 279-287.
[24] GONG B Q, SHI Y, SHA F, et al. Geodesic flow kernel for unsupervised domain adaptation[C]// 2012 Conference on Computer Vision and Pattern Recognition. Providence: IEEE, 2012: 2066-2073.
[25] SAENKO K, KULIS B, FRITZ M et al. Adapting visual category models to new domains[C]// 11th European Conference on Computer Vision, Heraklion. Greece: Springer, 2010: 213-226.
[26] WANG Q, BRECKON T P. Unsupervised domain adaptation via structured prediction based selective pseudo-labeling[C]//The Thirty-Fourth Conference on Artificial Intelligence. New York: AAAI, 2020: 6243-6250.
[27] NENE S A, NAYAR S K, MURASE H. Columbia object image library (COIL-100) [R]. New York: Columbia University, 1996.
[28] VENKATESWARA H, EUSEBIO J, CHAKRABORTY S, et al. Deep hashing network for unsupervised domain adaptation[C]// 2017 Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 5018-5027.
[29] LIU H, LONG M S, WANG J M, et al. Transferable adversarial training: a general approach to adapting deep classifiers[C] //2019 Proceedings of the 36th International Conference on Machine Learning. California: PMLR, 2019: 4013-4022.
[30] ZHU Y C, ZHUANG F Z, WANG J D, et al. Deep subdomain adaptation network for image classification [J]. Journal of IEEE Transactions on Neural Networks and Learning Systems, 2020, 32: 1713-1722.
[31] LI J J, CHEN E, DING Z M, et al. Maximum density divergence for domain adaptation [J]. Journal of IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43: 3918-3930.
[1] 曾安, 赖峻浩, 杨宝瑶, 潘丹. 基于多尺度卷积和注意力机制的病理图像分割网络[J]. 广东工业大学学报, 2024, 41(0): 0-.
[2] 陈鑫瑀, 朱鉴, 陈炳丰, 蔡瑞初. 一种基于邻居环境感知的主动领域自适应算法[J]. 广东工业大学学报, 0, (): 19-.
[3] 古慧敏, 肖燕珊, 刘波. 基于多核学习的单分类多示例学习算法[J]. 广东工业大学学报, 2024, 41(02): 101-107.
[4] 钟耿君, 李东. 基于通道分离机制的双分支点云处理网络[J]. 广东工业大学学报, 2023, 40(04): 18-23.
[5] 朱弘扬, 丁怡, 柴华金, 李升. 可拓逻辑和可拓思维对决策树算法的分析与优化[J]. 广东工业大学学报, 2019, 36(02): 31-36.
[6] 陈平华, 黄辉, 麦淼, 周宏虹. 结合ReliefF和互信息的多标签特征选择算法[J]. 广东工业大学学报, 2018, 35(05): 20-25,50.
[7] 饶东宁, 黄思宏. 基于THUCTC的金融语料情感分析模型优化[J]. 广东工业大学学报, 2018, 35(03): 37-42.
[8] 张巍, 黄健华, 刘冬宁, 滕少华, 刘子婷. 一种改进的结合评分和评论信息的推荐方法[J]. 广东工业大学学报, 2017, 34(06): 27-31,48.
[9] 陈炳丰, 郝志峰, 蔡瑞初, 温雯, 王丽娟, 黄浩, 蔡晓凤. 面向汽车评论的细粒度情感分析方法研究[J]. 广东工业大学学报, 2017, 34(03): 8-14.
[10] 杨积升, 章云, 李东. 点云目标检测残差投票网络[J]. 广东工业大学学报, 2022, 39(01): 56-62.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
版权所有 © 2017 《广东工业大学学报》编辑部
地址：广州市东风东路729号广东工业大学 邮编：510090
电话：020-37626653,020-37626139,020-37626396
邮箱：xbzrb@gdut.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发