基于核典型相关分析的多视图谱聚类算法  

Multi-view Spectral Clustering Algorithm Based on Kernel Canonical Correlation Analysis

作　　者：王春杰 石延新 何进荣 王文发[1] WANG Chunjie;SHI Yanxin;HE Jinrong;WANG Wenfa(School of Mathematics and Computer Science,Yan’an University,Yan’an 716000,China)

机构地区：[1]延安大学数学与计算机科学学院,陕西延安716000

出　　处：《延安大学学报（自然科学版）》2021年第4期48-53,共6页Journal of Yan'an University：Natural Science Edition

基　　金：国家自然科学基金项目(61902339);国家自然科学基金地区项目(618660389);陕西省自然科学基础研究计划资助项目(2021JM-418);延安市科技专项资助项目(2019-01,2019-13);延安大学博士科研启动项目(YDBK2019-06)。

摘　　要：提出并讨论了一种基于核典型相关谱聚类假设的多视图协同聚类方法。目的是寻求从多个表示中找到一致的解决方法,利用每个表示中的信息来提高经典聚类模型的性能。为了展示该方法的有效性,在6个具有不同大小和特征的图像数据集、文本数据集上进行了实验。实验结果充分证实了基于核典型相关分析的多视图谱聚类算法的性能优于经典的单视图K-means、K-means级联(CK)、多视图K-means(MK)、单视图谱聚类(SC)、SC级联(CSC)和多视图谱聚类(MSC),且具有更好的聚类准确度和鲁棒性。A Multi-view collaborative clustering method based on Kernel Canonical Correlation Spectral Clustering assumption is proposed and discussed to find a consistent solution from multiple representations and improve the performance of classical clustering systems by using the information in each representation.In order to demonstrate the effectiveness of the proposed method,experiments were carried out on six image data sets and text data sets with different sizes and features.The experimental results showed that the performance of Multi-view Spectral Clustering Algorithm based on Kernel Canonical Correlation Analysis is better than Single-view K-means,K-means Cascade(CK),Multi-view K-means(MK),Single-view Spectral Clustering(SC),Spectral Clustering Cascade(CSC),Multi-view Spectral Clustering(MSC)as it has better clustering accuracy and robustness.

关 键 词：核典型相关分析 多视图K-means 多视图谱聚类 核典型相关分析的多视图谱聚类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]