Co-consensus Multi-view Spectral Clustering

Multi-view learning has attracted wide attention because of the ability to integrate information from different views. For the issue of multi-view data fusion, a co-consensus multi-view spectral clustering method is proposed. The method adds two consensus constraints in the model of spectral clustering to utilize the feature relationship and the similarity relationship of different views’ spectral embedding matrices which enhances the consistency of multi views. Simultaneously, this method obtains closed solution of the consensus variables in the optimization process, which further improves the clustering performance. The experiment tests the convergence, parameter sensitivity and clustering performance of the proposed method in three real-world datasets. The experiment results show that this method has the best performance in multiple performance metrics compared with the existing methods, and the maximum improvement in clustering accuracy is more than 10%. The experiment proves the co-consensus method effectively improves the performance of multi-view spectral clustering algorithm.