Graph Regularized Structured Sparse Subspace Clustering

High-dimensional data presents a big challenge for the clustering problem, however, the high-dimensional data often lie in low-dimensional subspaces. So, subspace clustering has been widely researched. Sparse subspace clustering (SSC) is considered as the state-of-the-art method for subspace clustering, it has received an increasing amount of interest in recent years. In this paper, we propose a novel sparse subspace clustering method named graph regularized structured sparse subspace clustering (GS3C) to jointly analyze the data under a single clustering framework and with shared underlying sparse representations. Two convex regularizations are combined and used in the model to enable sparsity as well as to facilitate multi-task learning. We also introduced the graph regularization to improve stability and consistency. The effectiveness of the proposed algorithm is demonstrated through experiments on motion segmentation and face clustering.