中文摘要：

在现有的稀疏子空间聚类算法理论基础上提出一个改进的稀疏子空间聚类算法：迭代加权的稀疏子空间聚类。稀疏子空间聚类通过解决l1最小化算法并应用谱聚类把高维数据点聚类到不同的子空间,从而聚类数据。迭代加权的l1算法比传统的l1算法有更公平的惩罚值,平衡了数据数量级的影响。此算法应用到稀疏子空间聚类中,改进了传统稀疏子空间聚类对数据聚类的性能。仿真实验对Yale B人脸数据图像进行识别分类,得到了很好的聚类效果,证明了改进算法的优越性。

英文摘要：

Based on the existing theory of sparse subspace clustering algorithm,a modified sparse subspace clustering algorithm is put forward：iterative weighted sparse subspace clustering algorithm.In order to cluster data,sparse subspace clustering algorithm clusters high-dimensional data to different subspaces by solving minimization algorithm and applying spectral clustering.Iterative algorithm has more fair punishment value then the traditional algorithm,with balancing the influence of magnitude of data.The algorithm is applied to the sparse subspace clustering to improve the traditional sparse subspace clustering performance for data. Simulation experiment recognizing and classify Yale B face data image.The clustering effect is very good,proving the superiority of the improved algorithm.