Partially Shared Sparse Representation With Graph Regularizer For Multi-View Clustering

Recently, multi-view clustering has been widely studied as multi-view data can provide more useful information for improving the clustering. The performance of existing multi-view clustering algorithms is often closely related to the learned representation. Thus, how to learn an effective representation is a fundamental problem. Considering the consistency and the complementarity of the multi-view data, this paper develops a partially shared sparse representation method. To be specific, the learned sparse representation for each view is composed of two parts: one is shared by all views and the other is a view-specific part. In addition, the graph regularizer is added to the proposed model to capture the geometric structure of multi-view data for better clustering. To solve efficiently the proposed model, we design an alternating minimization algorithm based on the accelerated proximal gradient (APG) scheme. In the end, experiments on three real-world datasets are conducted to show the superiority of our approach over the state-of-the-art algorithms, including both negative data and positive data.