南京大学学报（自然科学版）2024，Vol.60Issue(5)：745-752,8.DOI:10.13232/j.cnki.jnju.2024.05.005
扩散策略增强的多视角图聚类
Diffusion strategy enhanced multi-view graph clustering
摘要
Abstract
As one of the important directions in clustering analysis,deep graph clustering is attracting more and more attention from academia and industry.Existing deep graph clustering methods tend to focus on the model-level,that is,by adjusting the module structure to achieve better performance.It means that existing methods usually ignore information enhancement at the data-level and rely solely on the original data.When there is noise or loss in the original data,the performance of the graph clustering methods that relies heavily on it will decline to varying degrees.To solve this problem,we propose a novel method that utilizes the diffusion strategy to enhance Multi-View Graph Clustering,called DMVGC.Specifically,DMVGC applies the idea of diffusion technology to graph data and promotes more effective edges based on the original graph structure to generate multiple views,thereby providing the model with a broader clustering view and richer information propagation.The experimental results show that this DMVGC method achieves performance improvement compared with existing methods without affecting the training speed.关键词
聚类分析/深度图聚类/数据增强/图扩散策略/多视角学习Key words
clustering analysis/deep graph clustering/data augmentation/graph diffusion strategy/multi-view learning分类
信息技术与安全科学引用本文复制引用
刘猛,梁科,孟令源,李昊,周思航,刘新旺..扩散策略增强的多视角图聚类[J].南京大学学报（自然科学版）,2024,60(5):745-752,8.基金项目
国家自然科学基金杰出青年科学基金(62325604) （62325604）
