Neighbor structure aware based cross-view consensus graph learning for incomplete multi-view clustering

作     者：Yang, Haifeng Li, Xingyi Chen, Yijing 

作者机构：Jiangsu Univ Sch Comp Sci & Commun Engn Zhenjiang Peoples R China Jiangsu Univ Sch Comp Sci Zhenjiang Peoples R China Jiangsu Univ Sch Comp Sci & Commun Engn Comp Technol Zhenjiang Peoples R China 

出 版 物：《NEUROCOMPUTING》 (Neurocomputing)

年 卷 期：2025年第624卷

核心收录：

学科分类：08[工学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：Thanks to the authors for contributing to the paper 

主　　题：Graph learning Incomplete multi-view clustering Cross-view representation Structure preservation 

摘      要：Multi-view clustering has gained significant attention due to its ability to fully leverage the relationships between different views. In the face of the incomplete observability of complex systems, incomplete multi-view clustering (IMVC) has become an important research direction in machine learning. Currently, graph-based methods have shown excellent performance in clustering research. However, there are still some shortcomings: first, some methods only focus on non-missing sample information;second, most only explore pairwise relationships between samples while neglecting connections between neighboring clusters;finally, many methods fail to simultaneously utilize global structural information and potential information between views. To address these issues, this paper proposes a neighbor-group structure-aware cross-view incomplete multi-view clustering method (NACG_IMVC), aimed at exploring the relationships between sample neighbor groups. This method constructs a high-quality consensus graph by leveraging intra-view local information and inter-view global structural information. Additionally, a novel indicator matrix is designed to account for the structure of missing data, thereby enhancing the quality of the consensus graph. Experimental results on multiple datasets indicate that the clustering performance of NACG_IMVC surpasses that of several advanced clustering methods.