Multi-view Clustering with Noisy Views

Abstract

In real-word applications of multi-view clustering, there may be noisy views, which hurt the performance of multi-view clustering. To address this issue, we propose a novel multi-view clustering method which assigns the noisy views with smaller weights to alleviate the negative effect of the noisy views for better clustering performance. Specifically, we simultaneously learn the kernel k-means clustering structures in each view, the latent consistent clustering structure between views and the weights of views. The noisy views are allowed to have worse kernel k-means performance while have consistent clustering structure by assigning smaller weights. To solve the corresponding optimization problem, we develop an efficient algorithm, which updates partial variables alternatively. Moreover, the convergence of the proposed algorithm is theoretically guaranteed. Comprehensive experiments both on conventional multi-kernel dataset and synthetic noisy multi-kernel dataset demonstrate the efficacy of the proposed method.