自适应融合相似图的多视图谱聚类算法  

Adaptive Multi-view Spectral Clustering Algorithm with Fusion Similarity Graph

作　　者：路梅 杨雨萱[1] LU Mei;YANG Yuxuan(Jinling Institute of Technology,Nanjing 211169,China;Soochow University,Suzhou 215006,China)

机构地区：[1]金陵科技学院软件工程学院,江苏南京211169 [2]江苏省网络空间安全工程实验室(苏州大学),江苏苏州215006

出　　处：《金陵科技学院学报》2024年第3期1-12,共12页Journal of Jinling Institute of Technology

基　　金：江苏省产学研合作项目(BY20230040);金陵科技学院高层次人才科研启动项目(jit-b-202045);国家自然科学基金面上项目(62177028);江苏省重点实验室开发项目(SDGC2238);金陵科技学院横向项目(jit-h-2023-143);教育部产学合作协同育人项目(231100461100144)。

摘　　要：针对传统多视图谱聚类算法中二步法导致的公共视图质量低、权重超参确定困难等问题,提出了一种新的自适应融合相似图的多视图谱聚类算法。该算法利用各视图的一致性和互补性,充分挖掘数据隐藏的结构信息,依据各视图对公共视图的贡献自适应分配权重。不同于传统多视图谱聚类算法的二步法策略,该算法由公共矩阵直接得到聚类结果。在6个真实数据集上的实验验证了该算法的有效性。In order to solve the problems of low quality of common view and the difficulty in determining the hyperparameters of weights,produced by the two-step method in traditional multi-view spectral clustering algorithms,this paper proposes a new algorithm called adaptive multi-view spectral clustering with fusion similarity graph(MSCFSG).The method utilizes the consistency and complementarity of each view to fully explore the hidden structural information of data,and adaptively assigns weights based on the contribution of each view to the common view.Unlike the two-step strategy of traditional multi-view spectral clustering algorithms,the algorithm directly obtains clustering results from a common matrix.The effectiveness of the algorithm has been validated through experiments on six real datasets.

关 键 词：谱聚类 多视图聚类 融合相似图 非负矩阵分解 自动加权技术 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]