Abstract:Traditional subspace clustering algorithms need to transform each sample into a vector form. Therefore, problems of high dimensionality and small size samples are caused, the natural structural information of each sample is ignored and the clustering information is missing. To overcome the drawbacks, the weighted block subspace clustering based on least square regression algorithm ( WB-LSR) is proposed. Firstly, each sample is divided into lots of blocks, and the corresponding affinity matrices of each block are obtained. Next, the weight of each affinity matrix is determined by mutual vote between affinity matrices. Finally, the weighted sum of affinity matrices is regarded as final affinity matrix. The experimental results on image datasets and motion segmentation video datasets show that the proposed method effectively improves clustering accuracy.

Weighted Block Subspace Clustering Based on Least Square Regression