南京大学学报（自然科学版）Issue(4)：482-493,12.DOI:10.13232/j.cnki.jnju.2014.04.013
基于属性最大间隔的子空间聚类
Maximum margin model for subspace clustering
摘要
Abstract
Subspace Clustering can effectively discover the relationship between clusters and the subspaces,and it can reduce the interference caused by data redundancy and unrelated features in high dimensional datasets.Existing Subspace Clustering algorithms focus on the detection of clusters in subspace,while the division of subspace is ignored.This paper proposed a Subspace Clustering method based on features maximum margin,and its main idea is that minimum information is lost during the divide of subspaces,so the results of subspace clustering are better. There are two works in this paper.Firstly,the obj ective function of the subspace division is stated,and it makes the dependence of different subspaces to be minimum.Secondly,Subspace Clustering algorithm Maximum Margin Subspace Clustering(MMSC)based on maximum margin is designed for Subspace Clustering ensemble.At last,UCI and NIPS2013 competition datasets are used for experiments and the results show that MMSC algorithm on most datasets performs better results than other Subspace Clustering algorithms.关键词
子空间聚类/最大间隔/最大间隔子空间聚类Key words
subspace clustering/maximum margin/maximum margin subspace clustering引用本文复制引用
刘波,王红军,成聪,杨燕..基于属性最大间隔的子空间聚类[J].南京大学学报（自然科学版）,2014,(4):482-493,12.基金项目
国家自然科学基金(61003142,61262058,61175047,61170111) （61003142,61262058,61175047,61170111）
