Using Genetic algorithm to enhance nonnegative matrix factorization initialization

Recently, there is a growing interest to improve Non-negative Matrix Factorization (NMF) performance by developing efficient initialization methods. The aim of this paper is to estimate initial values for NMF components using Genetic algorithms (GAs). As far as NMF methods suffer from lack of convexity, the proposed method, here called NMF_GA can find a near optimal solution to initialize the NMF components. The proposed method was applied to JAFFE facial expression dataset. Results achieved by GA-NMF were compared to vast variety of NMF initialization methods and the supremacy of the obtained results showed the effectiveness of our GA-NMF method.