Subspace Clustering with Block Diagonal Sparse Representation

Structured representation is of remarkable significance in subspace clustering. However, most the existing clustering algorithms resort to single-structured representation, which may fail fully capture essential characteristics data. To address this issue, a novel multi-structured algorithm called block diagonal sparse (BDSR) proposed paper. It takes both and structured representations into account obtain desired affinity matrix. The unified framework established by integrating prior original resulting optimization problem iteratively solved inexact augmented Lagrange multipliers (IALM). Extensive experiments on synthetic real-world datasets well demonstrate effectiveness efficiency against state-of-the-art algorithms.