Unsupervised Multi-View Clustering by Squeezing Hybrid Knowledge From Cross View and Each View

Multi-view clustering methods have been a focus in recent years because of their superiority performance. However, typical traditional multi-view algorithms still shortcomings some aspects, such as removal redundant information, utilization various views and fusion features. In view these problems, this paper proposes new method, low-rank subspace based on adaptive graph regularization. We construct two data matrix decomposition models into unified optimization model. framework, we address the significance common knowledge shared by cross unique each presenting sparse constraints matrix. To ensure that achieve effective representation performance original matrix, regularization unsupervised are also incorporated proposed model to preserve internal structural features data. Finally, method is compared with several state-of-the-art algorithms. Experimental results for five widely used benchmarks show our algorithm surpasses other clear margin.