Nonnegative Matrix Factorization using Class Label Information

Abstract

Nonnegative matrix factorization (NMF) has been a powerful tool for finding out parts-based, linear representations of nonnegative data samples. Nevertheless, NMF is an unsupervised algorithm, and it is not able to utilize the class label information. In this paper, the Nonnegative Matrix Factorization using Class Label Information (NMF-CLI) is proposed. It combines the class label information for factorization constraints. The proposed NMF-CLI method is investigated with one cost function and the corresponding update rules are given. Experiment results show the power of the proposed novel algorithm, by comparing to the state-of-the-art methods.