Coordinate Ranking Regularized Non-negative Matrix Factorization

Abstract

Non-negative Matrix Factorization (NMF) has become increasingly popular in many applications that require data mining techniques such as information retrieval, computer vision, and pattern recognition. NMF aims at approximating the original data matrix in a high dimensional space with the product of two non-negative matrices in a lower dimensional space. In many applications with high dimensional data such as text, data often have a global geometric structure, which typically may not be directly derived from the local information. But the existing literature of NMF completely ignores this problem. This paper proposes a novel matrix factorization algorithm called Coordinate Ranking regularized NMF (CR-NMF) in order to address this problem. The idea of the proposed algorithm is to combine NMF and manifold ranking to encode both local and global geometric structures of the data. Experimental results on two real-world datasets demonstrate the superiority of this algorithm.