SGM to solve NMF – Application to Hyperspectral Data

Laboratoire Lagrange, Université de Nice Sophia-Antipolis, Observatoire de la Côte d’Azur, CNRS, Nice, France

Abstract

This article deals with the problem of minimization of a general cost function under non-negativity and flux conservation constraints. The proposed algorithm is founded on the Split Gradient Method (SGM) adapted here to solve the Non Negative Matrix Factorization (NMF). We show that SGM can be easily regularized, allowing to introduce some physical constraints. Finally, to validate the algorithm, we propose an example of application to hyperspectral data unmixing.