Multi-view clustering based on a multimetric matrix fusion method

Abstract

Multi-view clustering(MVC) utilizes the consistency of multiple views to learn a consensus representation. However, the existing MVC methods usually use only a single metric to learn the graph matrix, which cannot fully reveal the real structure between complex samples and makes the clustering performance unsatisfactory. To solve this problem, we propose a novel method, i.e., multi-view clustering based on a multimetric matrix fusion method(MVC3MF). Specifically, we first concatenate the multi-view data into a joint representation. Then, we learn multiple kinds of distance metric matrices based on the joint representation. Third, we fuse the obtained multimetric matrices into an optimal metric matrix by using an adaptive weight method. Finally, we use the optimal metric matrix to learn the graph matrix, which is imposed by a rank constraint. To verify the superiority of our algorithm, we performed experiments on six datasets and the experimental results show that the clustering performance of MVC3MF is superior to that of some state-of-the-art MVC methods.