Latent Multi-view Subspace Clustering Based on Schatten-P Norm

In this paper, we aim at the research of rank minimization to find more accurate low-dimensional representations for multi-view subspace learning. The Schatten-p norm is utilized as the rank relaxation function for subspace learning to enhance its ability to recover the low rank matrices, and a multi-view subspace clustering algorithm via maximizing the original feature information is proposed under the assumption that each view is derived from a latent representation. With the Schatten-p norm, the proposed algorithm can improve the quality and robustness of the latent representations. The effectiveness of our method is validated through experiments on several benchmark datasets.