优化加权多视角K-means聚类算法
|
摘要
现存的多视角聚类算法能够充分利用多个视角的信息进行聚类,因而其聚类效果较单视角聚类算法更优,但是绝大多数多视角聚类算法在聚类过程中为各个视角赋予了同等的权重值,这对于划分不明确的视角,会严重影响聚类的最终结果。目前的加权K-means聚类算法在面对多视角聚类任务时,能解决上述权重的取值分配问题,但其权重在迭代过程中会出现除以零错误,造成相关视角的丢失。针对这个问题,提出了一种优化加权多视角K-means聚类算法(MKSC)。该算法给每个视角分配权重,利用加权策略有效地控制各个视角的重要程度,通过引入常数对每个视角的权重进行优化,使用K-means进行聚类。通过基于人工数据集和真实数据集的实验对该算法进行验证,实验结果表明该算法较已有的多视角聚类技术具有更好的聚类性能。
The existing multi-view clustering algorithm can make full use of multi-view information to cluster,so its effect is better than that of the single view clustering algorithm. However,in the clustering process,most of the multi-view clustering algorithms assign the same weight values for each view,which will seriously affect the final result of clustering. The current weighted K-means clustering algorithm can solve the above problem of weight assigning for the multi-view clustering tasks,but its weight will be divided by zero in the iteration process,which leads to the loss of related perspectives. For this,we propose an optimizing weighted multi-view K-means clustering algorithm(MKSC) which assigns weight for each view and uses the weighted strategy to effectively determine the importance of the various perspectives,optimizing the weight of each view by introducing a constant and with K-means to cluster. The algorithm is verified by experiments based on artificial data set and real dataset,results of which have shown that it has better clustering performance than the existing multi view clustering technology.
The existing multi-view clustering algorithm can make full use of multi-view information to cluster,so its effect is better than that of the single view clustering algorithm. However,in the clustering process,most of the multi-view clustering algorithms assign the same weight values for each view,which will seriously affect the final result of clustering. The current weighted K-means clustering algorithm can solve the above problem of weight assigning for the multi-view clustering tasks,but its weight will be divided by zero in the iteration process,which leads to the loss of related perspectives. For this,we propose an optimizing weighted multi-view K-means clustering algorithm(MKSC) which assigns weight for each view and uses the weighted strategy to effectively determine the importance of the various perspectives,optimizing the weight of each view by introducing a constant and with K-means to cluster. The algorithm is verified by experiments based on artificial data set and real dataset,results of which have shown that it has better clustering performance than the existing multi view clustering technology.
引文
[1] 熊子源,徐振海,张亮,等.基于聚类算法的最优子阵划分方法研究[J].电子学报,2011,39(11):2615-2621.
[2] 罗恩韬,王国军.大数据中一种基于语义特征阈值的层次聚类方法[J].电子与信息学报,2015,37(12):2795-2801.
[3] 刘卓,杨悦,张健沛,等.不确定度模型下数据流自适应网格密度聚类算法[J].计算机研究与发展,2014,51(11):2518-2527.
[4] MIYAHARA S,MIYAMOTO S.A family of algorithms using spectral clustering and DBSCAN[C]//IEEE international conference on granular computing.Noboribetsu,Japan:IEEE,2014:196-200.
[5] 邱保志,贺艳芳.多视角核K-means聚类算法的收敛性证明[J].郑州大学学报:理学版,2017,49(3):32-38.
[6] 蒋亦樟,邓赵红,王骏,等.熵加权多视角协同划分模糊聚类算法[J].软件学报,2014,25(10):2293-2311.
[7] 刘正,张国印,陈志远.基于特征加权和非负矩阵分解的多视角聚类算法[J].电子学报,2016,44(3):535-540.
[8] WANG Jidong,ZENG Huajun,CHEN Zheng,et al.ReCoM:rein-forcement clustering of multi-type interrelated data objects[C]//Proceedings of the 26th annual international ACM SIGIR conference on research and development in information retrieval.Toronto,Canada:ACM,2003:274-281.
[9] BICKEL S, SCHEFFER T. Multi-view clustering[C]//IEEE international conference on data mining.Brighton,UK:IEEE,2004:19-26.
[10] BICKEL S,SCHEFFER T.Estimation of mixture models using Co-EM[C]//Proceedings of the 16th European conference on machine learning.Porto,Portugal:Springer-Verlag,2005:35-46.
[11] CHAUDHURI K,KAKADE S M,LIVESCU K,et al.Multi-view clustering via canonical correlation analysis[C]//Proceedings of international conference on machine learning.New York,NY,USA:ACM,2009:129-136.
[12] TZORTZIS G F,LIKAS C L.The global kernel -means algorithm for clustering in feature space[J].IEEE Transactions on Neural Networks,2009,20(7):1181-1194.
[13] DENG Zhaohong,CHOI K S,CHUNG F L,et al.Enhanced soft subspace clustering integrating within-cluster and between-cluster information[J].Pattern Recognition,2010,43(3):767-781.
[14] CLEUZIOU G,EXBRAYAT M,MARTIN L,et al.CoFKM:a centralized method for multiple-view clustering[C]//Ninth IEEE international conference on data mining.Miami,FL,USA:IEEE,2009:752-757.
[15] GU Quanquan,ZHOU Jie.Learning the shared subspace for multi-task clustering and transductive transfer classification[C]//IEEE international conference on data mining.Miami,FL,USA:IEEE,2009:159-168.
[16] GU Quanquan,ZHOU Jie.Co-clustering on manifolds[C]//ACM SIGKDD international conference on knowledge discovery and data mining.Paris,France:ACM,2009:359-368.
[2] 罗恩韬,王国军.大数据中一种基于语义特征阈值的层次聚类方法[J].电子与信息学报,2015,37(12):2795-2801.
[3] 刘卓,杨悦,张健沛,等.不确定度模型下数据流自适应网格密度聚类算法[J].计算机研究与发展,2014,51(11):2518-2527.
[4] MIYAHARA S,MIYAMOTO S.A family of algorithms using spectral clustering and DBSCAN[C]//IEEE international conference on granular computing.Noboribetsu,Japan:IEEE,2014:196-200.
[5] 邱保志,贺艳芳.多视角核K-means聚类算法的收敛性证明[J].郑州大学学报:理学版,2017,49(3):32-38.
[6] 蒋亦樟,邓赵红,王骏,等.熵加权多视角协同划分模糊聚类算法[J].软件学报,2014,25(10):2293-2311.
[7] 刘正,张国印,陈志远.基于特征加权和非负矩阵分解的多视角聚类算法[J].电子学报,2016,44(3):535-540.
[8] WANG Jidong,ZENG Huajun,CHEN Zheng,et al.ReCoM:rein-forcement clustering of multi-type interrelated data objects[C]//Proceedings of the 26th annual international ACM SIGIR conference on research and development in information retrieval.Toronto,Canada:ACM,2003:274-281.
[9] BICKEL S, SCHEFFER T. Multi-view clustering[C]//IEEE international conference on data mining.Brighton,UK:IEEE,2004:19-26.
[10] BICKEL S,SCHEFFER T.Estimation of mixture models using Co-EM[C]//Proceedings of the 16th European conference on machine learning.Porto,Portugal:Springer-Verlag,2005:35-46.
[11] CHAUDHURI K,KAKADE S M,LIVESCU K,et al.Multi-view clustering via canonical correlation analysis[C]//Proceedings of international conference on machine learning.New York,NY,USA:ACM,2009:129-136.
[12] TZORTZIS G F,LIKAS C L.The global kernel -means algorithm for clustering in feature space[J].IEEE Transactions on Neural Networks,2009,20(7):1181-1194.
[13] DENG Zhaohong,CHOI K S,CHUNG F L,et al.Enhanced soft subspace clustering integrating within-cluster and between-cluster information[J].Pattern Recognition,2010,43(3):767-781.
[14] CLEUZIOU G,EXBRAYAT M,MARTIN L,et al.CoFKM:a centralized method for multiple-view clustering[C]//Ninth IEEE international conference on data mining.Miami,FL,USA:IEEE,2009:752-757.
[15] GU Quanquan,ZHOU Jie.Learning the shared subspace for multi-task clustering and transductive transfer classification[C]//IEEE international conference on data mining.Miami,FL,USA:IEEE,2009:159-168.
[16] GU Quanquan,ZHOU Jie.Co-clustering on manifolds[C]//ACM SIGKDD international conference on knowledge discovery and data mining.Paris,France:ACM,2009:359-368.