Consistent and Inconsistent Multi-view Clustering Based on Augmented View

Multi-view clustering plays an increasingly important role across various fields by fusing multiple forms of data. However, most current methods do not fully utilize the consistency and inconsistency information between multiple views. This paper proposes a new method. It connects each view to an enhanced matrix and then obtains the self-representation matrices for each view and the enhanced view. Each self-representation matrix is divided into two parts: consistent item and inconsistent items. The inconsistent items are stacked into a tensor, which utilizes the inconsistency information to establish connections between the multiple views, thereby enabling a more comprehensive exploration of the correlations among the views. At the same time, the role of consistency information should not be overlooked. The proposed method utilizes the ADMM algorithm for solving, and applies spectral clustering to obtain the final clustering results. Experimental results on six benchmark datasets confirm the effectiveness of the proposed model, demonstrating that its clustering performance is superior to existing multi-view clustering methods.