Block Diagonal Least Squares Regression for Subspace Clustering

Least squares regression (LSR) is an effective method that has been widely used for subspace clustering. Under the conditions of independent subspaces and noise-free data, coefficient matrices can satisfy enforced block diagonal (EBD) structures achieve good clustering results. More importantly, LSR produces closed solutions are easier to solve. However, with properties have solved using sensitive noise or corruption as they fragile easily destroyed. Moreover, when actual datasets, these cannot always guarantee satisfactory Considering representation excellent performance, idea constraints introduced into a new method, which named least (BDLSR), proposed. By regularizer, BDLSR effectively reinforce obtained improve performance. Our experiments several real datasets illustrated produced higher performance compared other algorithms.