An Efficient Non-negative Matrix Factorization with Its Application to Face Recognition

摘要

This paper attempts to develop a novel Non-negative Matrix Factorization (NMF) algorithm to improve traditional NMF approach. Based on gradient descent method, we appropriately choose a larger step-length than that of traditional NMF and obtain efficient NMF update rules with fast convergence rate and high performance. The step-length is determined by solving some inequalities, which are established according to the requirements on step-length and non-negativity constraints. The proposed algorithm is successfully applied to face recognition. The rates of both convergence and recognition are utilized to evaluate the effectiveness of our method. Compared with traditional NMF algorithm on ORL and FERET databases, experimental results demonstrate that the proposed NMF method has superior performance.