Separable Consistency and Diversity Feature Learning for Multi-View Clustering

Multi-view clustering has garnered growing attention due to its ability to learn consistent representation across different views in order to enhance clustering performance. The majority of current research concentrates on aligning the feature distribution of the potential space to capture view-common information, disregarding the conflict between consistency alignment and the reconstruction objective. In this paper, we propose a multi-view clustering method via Separable Consistency and Diversity Feature Learning (SCDFL) to address the aforementioned issue. The proposed method decouples potential feature into two components for learning consistency and diversity, respectively, and integrates these features for data reconstruction. The consistency and diversity feature are concatenated for spectral clustering. Extensive experiments have demonstrated that the proposed method achieves superior performance compared to several state-of-the-art methods.