Clustering complex networks and biological networks by nonnegative matrix factorization with various similarity measures

Identifying community structure in complex networks is closely related to clustering of data in other areas without an underlying network structure. In this paper, we propose a nonnegative matrix factorization (NMF)-based method for finding community structure. We first evaluate several similarity measures, such as diffusion kernel similarity, shortest path based similarity on several widely well-studied networks. Then, we apply NMF with diffusion kernel similarity to a large biological network, which demonstrates that our method can find biologically meaningful functional modules. Comparison with other algorithms also indicates the good performance of our method.