Clean and robust multi-level subspace representations learning for deep multi-view subspace clustering

Kaiqiang Xu, Kewei Tang, Zhixun Su, Hongchen Tan

Abstract: Highlights•We propose a new method to handle the nonlinear multi-view subspace clustering task.•We use multiple self-express layers to learn multi-level subspace representations.•We use robust principal component analysis to clean up the learned representations.•We introduce the common layer to learn the shared common subspace representation.•Our method has better clustering performance than other state-of-the-art methods.