Multi⁃view clustering aims to exploit the consensus and complementarity across different views,yet existing methods face two major bottlenecks: first,global topology modeling relies on predefined similarity measures and fixed neighborhoods,making it difficult to adapt to complex data distributions; second,sample⁃level view quality varies significantly,and static weighting strategies fail to characterize fine⁃grained reliability changes,particularly lacking robustness in missing data scenarios. To this end,we propose a multi⁃view clustering framework based on Dynamic Anchors and quality⁃aware Mixture⁃of⁃Experts (DAMC⁃MoE). First,a learnable dynamic anchor mechanism replaces traditional predefined similarity measures,achieving end⁃to⁃end deep coupling between topological structure modeling and feature representation learning. Building on this,a quality⁃aware mixture⁃of⁃experts module is introduced,which generates quality tokens from sample⁃level completeness and signal⁃to⁃noise ratio to guide the gating mechanism for adaptive routing,realizing a paradigm shift from conventional view⁃level weighting to sample⁃level fine⁃grained perceptual fusion. Finally,a three⁃level contrastive learning mechanism is constructed to jointly reinforce semantic alignment from inter⁃view,intra⁃view,and local⁃global perspectives. In comprehensive comparative experiments on 5 benchmark datasets against 11 state⁃of⁃the⁃art algorithms,DAMC⁃MoE demonstrates superior clustering performance. Friedman test results further indicate that DAMC⁃MoE achieves significantly higher average rankings across three clustering evaluation metrics compared to all baseline methods.
其次,设计了质量感知的混合专家融合模块.该模块摒弃传统的静态视图级加权策略,引入混合专家(Mixture of Experts,MoE)架构[25-26],通过专门的质量感知子网从完整度和信噪比两个维度动态评估每个样本的视图质量.随后,生成的“质量令牌”被注入门控网络,使门控机制能在样本层面动态分配专家激活权重,从而自动抑制高噪或缺失视图的干扰,同时增强高质量视图的贡献,实现“去伪存真”.此外,多专家架构支持模型从不同子空间组合特征,显著提升了表征能力.
最后,采用三层增强对比学习策略.近年来,对比学习在无监督表示学习中展现出强大的语义对齐能力,例如,Yan et al[27]的GCFAggMVC算法通过结构引导的对比学习对齐共识表示与视图特定表示.Cui et al[28]的DCMVC算法采用双对比驱动策略,通过动态聚类扩散和可靠邻居引导来实现类间分离与类内紧凑.为了进一步挖掘动态锚点与融合特征的语义潜力,本文设计了包含视图间一致性、视图内结构鲁棒性及局部⁃全局协同对齐的三重约束目标,这种多层次监督信号促使模型在对齐语义空间的同时,有效保留视图内细粒度几何结构,进一步增强复杂场景下的聚类性能.
图结构因其能显式建模样本间的拓扑关系,已成为多视图聚类的主流技术路线.早期工作主要依赖预定义的相似度函数(如欧氏距离、余弦相似度)构建固定图结构.例如,Kumar et al[10]提出共正则化谱聚类,需要先为每个视图计算固定的核矩阵(相似度矩阵),再通过正则化视图间的谱嵌入来保证聚类一致性.类似地,Zhan et al[13]的MVGL算法通过优化拉普拉斯矩阵的秩来改善聚类性能,但初始图结构仍基于预定义相似度函数,无法摆脱固定图结构的依赖.Nie et al[29]的AMGL引入自动学习图权重的机制,但优化过程仍基于预定义图结构,难以完全自适应数据分布的内在变化.随着深度学习的兴起,基于深度图聚类的方法应运而生.如SGCMC[30],DCRN[31],ASGCN[32],MGCN⁃FN[33]等工作,利用深度神经网络在潜在空间中提取高阶语义特征并构建图结构,显著提升了特征的抽象表达能力,但这些方法在图结构与聚类目标的协同优化方面仍有改进空间.其一,图结构的初始化往往基于预定义规则,尽管部分方法引入注意力机制或图扰动策略进行后续优化,但初始拓扑的构建仍依赖启发式假设,在面对复杂数据分布时可能受到初始化质量的影响.其二,图学习与聚类任务的整合深度有待进一步增强.虽然部分方法尝试通过伪标签反馈(如SGCMC)或双层关联约束(如DCRN)实现了一定程度的协同,但图拓扑的优化与聚类损失的传导路径往往存在间接性,使图结构难以在整个训练过程中持续适配聚类目标的动态变化.
混合专家(MoE)模型近年来在自然语言处理(Natural Language Processing,NLP)和计算机视觉(Computer Vision,CV)领域展现出强大的条件计算能力[38].典型代表包括NLP领域的DeepSeek⁃MoE,通过细粒度专家分割与共享专家隔离的稀疏激活策略实现大规模语言模型的高效训练与推理[39];CV领域的视觉Transformer稀疏结构(V⁃MoE)通过批处理优先路由,在保持精度的同时可以大幅降低计算量[40].其核心均为根据输入内容动态选择并激活最合适的子网络.受此启发,近期研究开始尝试将MoE架构迁移至多视图聚类任务.例如,Du et al[41]提出一种用于多视图K⁃Means的双重MoE框架,通过自适应校准簇级别的视图权重并划分区域级子空间以平衡视图贡献.Zhang et al[42]将MoE作为深度表征学习器,利用协作专家网络和平衡约束来增强多视图特征的多样性与互补性.然而,上述方法主要关注特征空间的多样性扩充或簇级的权重分配,往往将门控机制视为隐式的优化过程,缺乏对数据内源性质量(如视图完整度和信噪比)的显式感知,特别是在面对严重的视图缺失时,仅仅依赖隐式路由可能导致次优的融合策略.
AlSaadR, Abd⁃AlrazaqA, BoughorbelS,et al. Multimodal large language models in health care:Applications,challenges,and future outlook. Journal of Medical Internet Research,2024,26:e59505.
[2]
SerraA, GaldiP, TagliaferriR. Chapter 13⁃multi⁃view learning in biomedical applications∥Kozma R,Alippi C,Choe Y,et al. Artificial Intelligence in the Age of Neural Networks and Brain Computing. The 2nd Edition.New York:Academic Press,2019:265-280.
[3]
XiaL H, XuY, HuangC,et al. Graph meta network for multi⁃behavior recommendation∥Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York,NY,USA:Association for Computing Machinery,2021:757-766.
[4]
ChenW H, HsuC C, LaiY A,et al. Attribute⁃aware recommender system based on collaborative filtering:Survey and classification. Frontiers in Big Data,2020(2):49.
[5]
ChaoG Q, SunS L, BiJ B. A survey on multi⁃view clustering. IEEE Transactions on Artificial Intelligence,2021,2(2):146-168.
[6]
AnandS K, KumarS. Experimental comparisons of clustering approaches for data representation. ACM Computing Surveys,2022,55(3):1-33.
[7]
YangY, WangH. Multi⁃view clustering:A survey. Big Data Mining and Analytics,2018,1(2):83-107.
[8]
LiangW X, ZhangY H, KwonY,et al. Mind the gap:Understanding the modality gap in multi⁃modal contrastive representation learning∥Proceedings of the 36th International Conference on Neural Information Processing Systems. Red Hook,NY,USA:Curran Associates Inc.,2022:17612-17625.
[9]
NieF P, TianL, LiX L. Multiview clustering via adaptively weighted Procrustes∥Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. New York,NY,USA:Association for Computing Machinery,2018:2022-2030.
[10]
KumarA, RaiP, DauméH. Co⁃regularized multi⁃view spectral clustering∥Proceedings of the 25th International Conference on Neural Information Processing Systems. Red Hook,NY,USA:Curran Associates Inc.,2011:1413-1421.
[11]
AlshammariM, StavrakakisJ, TakatsukaM. Refining a k⁃nearest neighbor graph for a compu⁃tationally efficient spectral clustering. Pattern Recognition,2021,114:107869.
[12]
LiuY F, ChengH, CuiJ T. Revisiting k⁃nearest neighbor graph construction on high⁃dimensional data:Experiments and analyses. https://arxiv.org/abs/2112.02234,2021-12-04.
[13]
ZhanK, ZhangC Q, GuanJ P,et al. Graph learning for multiview clustering. IEEE Transactions on Cybernetics,2018,48(10):2887-2895.
[14]
ZhaoW H, LiG F, YangH Z,et al. Embedded feature selection on graph⁃based multi⁃view clustering∥Proceedings of the 38th AAAI Conference on Artificial Intelligence,the 36th Conference on Innovative Applications of Artificial Intelligence,the 14th Symposium on Educational Advances in Artificial Intelligence. Menlo Park,CA,USA:AAAI Press,2024:17016-17023.
[15]
WenJ, WuZ H, ZhangZ,et al. Structural deep incomplete multi⁃view clustering network∥Proceedings of the 30th ACM International Conference on Information & Knowledge Management. New York,NY,USA:Association for Computing Machinery,2021:3538-3542.
[16]
HaoW Y, PangS M, BaiX X,et al. Tensor⁃based incomplete multi⁃view clustering with low⁃rank data reconstruction and consistency guidance. IEEE Transactions on Circuits and Systems for Video Technology,2023,33(12):7156-7169.
ChenX L, CaiD. Large scale spectral clustering with landmark⁃based representation. Proceedings of the AAAI Conference on Artificial Intelligence,2011,25(1):313-318.
[23]
KangZ, ZhouW T, ZhaoZ T,et al. Large⁃scale multi⁃view subspace clustering in linear time. Proceedings of the AAAI Conference on Artificial Intelligence,2020,34(4):4412-4419.
[24]
YangB, ZhangX T, LiZ H,et al. Efficient multi⁃view K⁃means clustering with multiple anchor graphs. IEEE Transactions on Knowledge and Data Engineering,2023,35(7):6887-6900.
[25]
JacobsR A, JordanM I, NowlanS J,et al. Adaptive mixtures of local experts. Neural Computation,1991,3(1):79-87.
[26]
YukselS E, WilsonJ N, GaderP D. Twenty years of mixture of experts. IEEE Transactions on Neural Networks and Learning Systems,2012,23(8):1177-1193.
[27]
YanW Q, ZhangY Y, LvC L,et al. GCFAgg:Global and cross⁃view feature aggregation for multi⁃view clustering∥2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Vancouver,Canada:IEEE,2023:19863-19872.
[28]
CuiJ R, LiY T, HuangH,et al. Dual contrast⁃driven deep multi⁃view clustering. IEEE Transactions on Image Processing,2024,33:4753-4764.
[29]
NieF P, LiJ, LiX L. Parameter⁃free auto⁃weighted multiple graph learning:A framework for multiview clustering and semi⁃supervised classification∥Proceedings of the 25th International Joint Conference on Artificial Intelligence. Menlo Park,CA,USA:AAAI Press,2016:1881-1887.
[30]
XiaW, WangQ Q, GaoQ X,et al. Self⁃supervised graph convolutional network for multi⁃view clustering. IEEE Transactions on Multimedia,2022,24:3182-3192.
[31]
XiaoS X, DuS D, ChenZ L,et al. Dual fusion⁃propagation graph neural network for multi⁃view clustering. IEEE Transactions on Multimedia,2023,25:9203-9215.
PengG W, DornaikaF, CharafeddineJ. Multi⁃view learning with graph convolution networks adopting diverse graphs and genuine deep feature fusion. Artificial Intelligence Review,2025,58(9):290.
[34]
BaltrusaitisT, AhujaC, MorencyL P. Multimodal machine learning:A survey and taxonomy. IEEE Transactions on Pattern Analysis and Machine Intelligence,2019,41(2):423-443.
[35]
VaswaniA, ShazeerN, ParmarN,et al. Attention is all you need∥Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook,NY,USA:Curran Associates Inc.,2017:6000-6010.
[36]
LiuX W, ZhuX Z, LiM M,et al. Efficient and effective incomplete multi⁃view clustering. Proceedings of the AAAI Conference on Artificial Intelligence,2019,33(1):4392-4399.
[37]
YangM X, LiY F, HuP,et al. Robust multi⁃view clustering with incomplete information. IEEE Transactions on Pattern Analysis and Machine Intelligence,2023,45(1):1055-1069.
[38]
FedusW, ZophB, ShazeerN. Switch transformers:Scaling to trillion parameter models with simple and efficient sparsity. https://arxiv.org/abs/2101.03961,2022-06-16.
[39]
DaiD M, DengC Q, ZhaoC G,et al. DeepSeekMoE:Towards ultimate expert specia⁃lization in mixture⁃of⁃experts language models∥Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics. Volume 1:Long Papers. Stroudsburg:Association for Computing Linguistics,2024:1280-1297.
[40]
RiquelmeC, PuigcerverJ, MustafaB,et al. Scaling vision with sparse mixture of experts∥Proceedings of the 35th International Conference on Neural Information Processing Systems. Red Hook,NY,USA:Curran Associates Inc.,2021:8583-8595.
[41]
DuL, ShiY K, ChenY,et al. A dual mixture⁃of⁃experts framework for multi⁃view K⁃means clustering with view balance and regional sparsity. Information Fusion,2026,125:103449.
[42]
ZhangY H, CaiJ Y, WuZ H,et al. Mixture of experts as representation learner for deep multi⁃view clustering∥Proceedings of the 39th AAAI Conference on Artificial Intelligence,the 37th Conference on Innovative Applications of Artificial Intelligence,the 15th Symposium on Educational Advances in Artificial Intelligence. Menlo Park,CA,USA:AAAI Press,2025:22704-22713.
[43]
PengX, HuangZ Y, LvJ C,et al. COMIC:Multi⁃view clustering without parameter selection∥Proceedings of the 36th International Conference on Machine Learning. New York,NY,USA:PMLR,2019:5092-5101.
[44]
ChenM S, LinJ Q, LiX L,et al. Representation learning in multi⁃view clustering:A literature review. Data Science and Engineering,2022,7(3):225-241.
[45]
LiF F, FergusR, PeronaP. Learning generative visual models from few training examples:An incremental Bayesian approach tested on 101 object categories∥2004 Conference on Computer Vision and Pattern Recognition Workshop. Washington DC,USA:IEEE,2004:178.
[46]
SunM J, ZhangP, WangS W,et al. Scalable multi⁃view subspace clustering with unified anchors∥Proceedings of the 29th ACM International Conference on Multimedia. New York,NY,USA:Association for Computing Machinery,2021:3528-3536.
[47]
HuangD, WangC D, LaiJ H. Fast multi⁃view clustering via ensembles:Towards scalability,superiority,and simplicity. IEEE Transactions on Knowledge and Data Engineering,2023,35(11):11388-11402.
[48]
WanX H, LiuJ Y, YuH,et al. Contrastive continual multiview clustering with filtered structural fusion. IEEE Transactions on Neural Networks and Learning Systems,2025,36(8):14829-14842.
[49]
TrostenD J, LokseS, JenssenR,et al. Reconsidering representation alignment for multi⁃view clustering∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville,TN,USA:IEEE,2021:1255-1265.
[50]
LinY J, GouY B, LiuX T,et al. Dual contrastive prediction for incomplete multi⁃view representation learning. IEEE Transactions on Pattern Analysis and Machine Intelligence,2023,45(4):4447-4461.
[51]
XuJ, TangH Y, RenY Z,et al. Multi⁃level feature learning for contrastive multi⁃view clustering∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans,LA,USA:IEEE,2022:16051-16060.
[52]
BianJ T, XieX H, LaiJ H,et al. Multi⁃view contrastive clustering via integrating graph aggregation and confidence enhancement. Information Fusion,2024,108:102393.
[53]
YanW Q, YangT Y, TangC. Self⁃supervised semantic soft label learning network for deep multi⁃view clustering. IEEE Transactions on Multimedia,2025,27:4971-4983.