摘要
提出了一种新的面向图的一致性多视角稀疏聚类框架,该方法先将多视角数据分解为一致性与不一致性部分;然后采用相似性度量方法与KNN(K-nearest neighbor)算法对多视角数据进行分解与融合;再运用稀疏表示学习多视角图的一致性相似矩阵,进而通过谱聚类获取聚类结果。最后,设计并实现了一种交替迭代优化算法求解目标函数,并在八个多视角数据集上通过对比实验验证了该方法的有效性。
This paper proposed a new graph-oriented consistent multi-view sparse clustering framework.This method firstly decomposed multi-view data into consistency and inconsistency parts,used similarity measurement method and KNN(K-nearest neighbor)algorithm to decompose and fuse multi-view data,and then used sparse representation to learn the consistent simila-rity matrix of multi-view graphs.Finally,it obtained the clustering results through spectral clustering.In addition,this paper designed and implemented an alternate iterative optimization algorithm to solve the objective function.It verifies the effectiveness of the method through comparative experiments on eight multi-view datasets.
作者
刘瑜童
滕少华
张巍
Liu Yutong;Teng Shaohua;Zhang Wei(School of Computers,Guangdong University of Technology,Guangzhou 510006,China)
出处
《计算机应用研究》
CSCD
北大核心
2021年第8期2315-2320,共6页
Application Research of Computers
基金
国家自然科学基金资助项目(61972102)
广东省重点领域研发计划资助项目(2020B010166006)
广东省教育厅资助项目(粤教高函〔2018〕179号,粤教高函〔2018〕1号)
广州市科技计划资助项目(201903010107,201802030011,201802010026,201802010042,201604046017)。
关键词
多视角聚类
稀疏表示
图融合
一致性
相似性
multi-view clustering
sparse representation
graph fusion
consistency
similarity
