多视图融合的部分多标签学习

doi:10.3778/j.issn.1002-8331.2409-0217

摘要/Abstract

摘要： 部分多标签学习（partial multi-label learning，PML）旨在训练实例的标签集中同时处理真实标签和噪声标签。当前大多数PML只考虑单视图场景，却忽略了现实中多视图场景的情况。虽然部分研究利用了多视图特征信息，但多依赖于获取多视图的子空间，对多视图特征信息的学习不够充分。针对以上挑战，提出了一种新颖的多视图融合的PML预测模型PMLMF（partial multi-label learning based on multi-view fusion）。利用矩阵非负分解获得共享子空间，以捕获多视图的共享信息，并将其与原始的多视图数据集相融合，构建新的多视图数据集。利用低秩表示获得标签相关性系数矩阵，从而有效去除噪声并恢复相关标签。将模型推广到非线性版本，以有效地处理线性不可分割的问题。在7个多视图部分多标签数据集中进行了大量的实验，充分验证了该方法的有效性。

关键词: 部分多标签学习, 多视图融合, 标签相关性, 非线性

Abstract: Partial multi-label learning (PML) aims to process both real and noisy labels in the label set of training examples. Most current PMLs only consider single-view scenarios while ignoring the reality of multi-view scenarios. Although some studies have utilized multi-view feature information, they mostly rely on acquiring multi-view subspaces, and the learning of multi-view feature information is not sufficient. To address the above challenges, this paper proposes a novel PML prediction model PMLMF (partial multi-label learning based on multi-view fusion). Firstly, matrix non-negative decomposition is utilized to obtain the shared subspace in order to capture the shared information of multi-views, and it is fused with the original multi-view dataset to construct a new multi-view dataset. Secondly, the label correlation coefficient matrix is obtained using the low-rank representation so as to effectively remove the noise and recover the relevant labels. Finally, the model is generalized to a nonlinear version to effectively deal with the linear inseparable problem. A large number of experiments are conducted on seven multi-view partially multi-labeled datasets to fully validate the effectiveness of the method.

Key words: partial multi-label learning, multi-view fusion, label correlation, nonlinear

杨中柳, 殷俊. 多视图融合的部分多标签学习[J]. 计算机工程与应用, 2025, 61(24): 154-165.

YANG Zhongliu, YIN Jun. Partial Multi-Label Learning Based on Multi-View Fusion[J]. Computer Engineering and Applications, 2025, 61(24): 154-165.

参考文献

[1] XIE M K, HUANG S J. Partial multi-label learning[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2018: 4302-4309.
[2] YU G X, CHEN X, DOMENICONI C, et al. Feature-induced partial multi-label learning[C]//Proceedings of the 2018 IEEE International Conference on Data Mining. Piscataway: IEEE, 2018: 1398-1403.
[3] ZHANG Y S, WU J, CAI Z H, et al. Multi-view multi-label lea-rning with sparse feature selection for image annotation[J]. IEEE Transactions on Multimedia, 2020, 22(11): 2844-2857.
[4] TAN Q Y, YU G X, WANG J, et al. Individuality- and commonality-based multiview multilabel learning[J]. IEEE Transa-ctions on Cybernetics, 2021, 51(3): 1716-1727.
[5] CHEN Z S, WU X, CHEN Q G, et al. Multi-view partial multilabel learning with graph-based disambiguation[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2020: 3553-3560.
[6] WU J H, WU X, CHEN Q G, et al. Feature-induced manifold disambiguation for multi-view partial multi-label learning[C]//Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. New York: ACM, 2020: 557-565.
[7] XU N, WU Y D, QIAO C Y, et al. Multi-view partial multi-label learning via graph-fusion-based label enhancement[J]. IEEE Transactions on Knowledge and Data Engineering, 2023, 35(11): 11656-11667.
[8] ZHU Y, KWOK J T, ZHOU Z H. Multi-label learning with global and local label correlation[J]. IEEE Transactions on Knowledge and Data Engineering, 2018, 30(6): 1081-1094.
[9] XU L L, WANG Z, SHEN Z F, et al. Learning low-rank label correlations for multi-label classification with missing labels[C]//Proceedings of the 2014 IEEE International Conference on Data Mining. Piscataway: IEEE, 2015: 1067-1072.
[10] ZHAO D W, GAO Q W, LU Y X, et al. Non-aligned multi-view multi-label classification via learning view-specific labels[J]. IEEE Transactions on Multimedia, 2023, 25: 7235-7247.
[11] ZHAO D W, GAO Q W, LU Y X, et al. Consistency and diversity neural network multi-view multi-label learning[J]. Knowledge-Based Systems, 2021, 218: 106841.
[12] LI Z W, LYU G Y, FENG S H. Partial multi-label learning via multi-subspace representation[C]//Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence, 2020: 2612-2618.
[13] ZHANG M L, FANG J P. Partial multi-label learning via credible label elicitation[C]//Proceedings of the IEEE Transactions on Pattern Analysis and Machine Intelligence. Piscataway: IEEE, 2020: 3587-3599.
[14] YU T T, YU G X, WANG J, et al. Partial multi-label learning using label compression[C]//Proceedings of the 2020 IEEE International Conference on Data Mining. Piscataway: IEEE, 2021: 761-770.
[15] WANG H, YANG Y, LIU B, et al. A study of graph-based system for multi-view clustering[J]. Knowledge-Based Systems, 2019, 163: 1009-1019.
[16] YIN J, WU H W, SUN S L. Effective sample pairs based contrastive learning for clustering[J]. Information Fusion, 2023, 99: 101899.
[17] ZHANG X B, YANG Y, LI T R, et al. CMC: a consensus multi-view clustering model for predicting Alzheimer’s disease progression[J]. Computer Methods and Programs in Biomedicine, 2021, 199: 105895.
[18] ZHAO D W, GAO Q W, LU Y X, et al. Two-step multi-view and multi-label learning with missing label via subspace learning[J]. Applied Soft Computing, 2021, 102: 107120.
[19] ZHONG Q Y, LYU G Y, YANG Z. Align while fusion: a generalized nonaligned multiview multilabel classification method[J]. IEEE Transactions on Neural Networks and Learning Systems, 2025, 36(4): 7627-7636.
[20] TAN Q Y, YU G X, DOMENICONI C, et al. Incomplete multi-view weak-label learning[C]//Proceedings of the 27th International Joint Conference on Artificial Intelligence. New York: ACM, 2018: 2703-2709.
[21] SUN L J, FENG S H, LIU J, et al. Global-local label correlation for partial multi-label learning[J]. IEEE Transactions on Multimedia, 2022, 24: 581-593.
[22] ZHANG M L, ZHOU Z H. ML-KNN: a lazy learning app-roach to multi-label learning[J]. Pattern Recognition, 2007, 40(7): 2038-2048.
[23] ZHANG Y Q, SHI D M, GAO J B, et al. Low-rank-sparse subspace representation for robust regression[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 2972-2981.
[24] LIN Z C, LIU R S, SU Z X. Linearized alternating direction method with adaptive penalty for low-rank representation[C]//Proceedings of the 25th International Conference on Neural Information Processing Systems. New York: ACM, 2011: 612-620.
[25] CAI J F, CANDèS E J, SHEN Z W. A singular value thresholding algorithm for matrix completion[J]. SIAM Journal on Optimization, 2010, 20(4): 1956-1982.
[26] ZHANG M L, ZHOU Z H. A review on multi-label learning algorithms[J]. IEEE Transactions on Knowledge and Data Engineering, 2014, 26(8): 1819-1837.
[27] XIE M K, HUANG S J. Partial multi-label learning with noisy label identification[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(7): 3676-3687.
[28] ZOU Y Z, HU X G, LI P P, et al. Learning shared and non-redundant label-specific features for partial multi-label classification[J]. Information Sciences, 2024, 656: 119917.
[29] CHENG Y S, XU Y T, GE W X. Multi-view multi-label learning for label-specific features via glocal shared subspace learning[J]. Applied Intelligence, 2024, 54(21): 11054-11067.
[30] LYU G Y, FENG S H, LI Y D. Noisy label tolerance: a new perspective of partial multi-label learning[J]. Information Sciences, 2021, 543: 454-466.
[31] DEM?AR J. Statistical comparisons of classifiers over multiple data sets[J]. Journal of Machine Learning Research, 2006, 7: 1-30.