Graph Learning For Multiview Clustering

Clustering is a hot topic in recently years.Clustering problem is to aggregates the disorder data into different clusters.And the similar data belong to same cluster.Nowadays,graph-based method is increasingly used in clustering.Graph-based clustering generates the similarity matrix from original data and obtains clustering results by the similarity of data points.But the early methods rely on the per-constructed graph.A fixed graph is structured by Gaussian kernel and generates the clustering indicator to obtain final results.The separation of graph and clustering seriously affect the clustering performance.To address this problem,the adaptive graph clustering method rises gradually.Adaptive graph clustering learns the graph adaptively instead of constructing graph beforehand.According to the regularization constraint,a better graph and clustering indicator can be obtained.However,above method only consider the single view graph model.The single view of data can not reflect all around of the data and can not stand for the real data.Hence,we proposed the multiview clustering with adaptive graph algorithm.Our approach learn the different initial graphs from data points of different views adaptively,and the initial grpahs are further optimized with a rank constraint on the Laplacian matrix.Then,these optimized graphs are integrated into a global graph with a well-designed optimization procedure.The global graph is learned by the optimaization procedure with the same rank constraint on its Lpalcian matrix.Because of the rank constraint,the cluster indicator is obtained directly by the global graph without performing any graph cut technique and the k-means clustering.Experiments are conducted on several benchmark datasets to verify the effectiveness and superiority of the proposed multiview clustering with adaptive graph algorithm comparing to the state-of-the-art methods.