Robust Informed Split Gradient NMF Using Alpha Beta-Divergence For Source Apportionment
Robert Chreiky, Université du Littoral Côte d'Opale, LISIC, Calais
Gilles Delmaire, Université du Littoral Côte d'Opale, LISIC, Calais
Clément Dorffer, Université du Littoral Côte d'Opale, LISIC, Calais
Matthieu Puigt, Université du Littoral Côte d'Opale, LISIC, Calais
Gilles Roussel, Université du Littoral Côte d'Opale, LISIC, Calais
Antoine Abche, University of Balamand

Abstract:
Source apportionment is a very challenging topic for which non-negative source separation is well-suited. Recently, we proposed several informed Non-negative Matrix Factorization (NMF) for which some expert knowledge was introduced. These methods were all dealing with some set values of one factor together with the row sum-to-one property by either processing each constraint alternatingly or using a new pa- rameterization which involves all of them. However, this last method was sensitive to the presence of outliers. In this pa- per, we thus propose a new robust informed Split Gradient NMF method which is based on a weighted alpha-beta divergence cost function. Experiments conducted for several input SNR with and without outliers on simulated mixtures of particulate matter sources show the relevance of the new approach.