Low-rank kernel consistent multi-view subspace clustering

Multi-view subspace clustering aims to enhance data analysis by integrating information from diverse perspectives. However, existing methods often fail to simultaneously leverage both consistent and complementary information across different views, and most methods ignore the nonlinear structure in the data. In this study, we propose a novel low-rank kernel consistent multi-view subspace clustering algorithm named LRKCMS. Specifically, we factorize the coefficient matrices into three distinct factors to maintain rank consistency across different views. Concurrently, we explore the self-representative coefficient matrices of each view to fully harness their complementary information. To mitigate the influence of noise and outliers in the underlying data, we apply the Schatten p-norm to regularize the low-rank coefficient matrices. Additionally, to fully exploit the intrinsic nonlinear characteristics of the data, we introduce a kernel mapping term that adaptively solves the low-rank kernel mapping, ensuring that the data in the resulting kernel feature space are both low-rank and self-expressive. We evaluate the performance of LRKCMS on 10 benchmark datasets and compare it against 12 state-of-the-art methods. The results demonstrate that LRKCMS achieves superior clustering performance compared to these advanced multi-view clustering methods. The source code is available at https://github.com/wzhangwhu/LRKCMS.