南京师大学报（自然科学版）Issue(3)：21-28,8.
基于局部和全局信息的正则化迭代聚类
Iterative Clustering with Local and Global Regularization
摘要
Abstract
Clustering is an efficient method of data analysis,K-means method is one of the most popular algorithms. The algorithm only works when the cluster of data is convex. Spectral clustering avoids the problems of K-means method, however,parameters settings in similarity calculation, complex calculation and storage complexity all constraint the effectiveness of spectral clustering. In this paper,Iterative clustering with local and global regularization is proposed. In this method,we conduct a cluster with a part of data,and then we add a small amount of remaining data gradually to the iterative calculation. The proposed method has the advantages of traditional spectral clustering,exploring both the local and global regularization,and achieve the effective clustering for Large-scale data by an iteration method. Experimental results on several data sets show the greater performance on the method.关键词
凸形/谱聚类/局部正则化/全局正则化/迭代Key words
convex/spectral clustering/local regularization/global regularization/iterative分类
信息技术与安全科学引用本文复制引用
许小龙,王士同..基于局部和全局信息的正则化迭代聚类[J].南京师大学报（自然科学版）,2014,(3):21-28,8.基金项目
江苏省自然科学基金(BK2011417) （BK2011417）
