Joint majorization-Minimization for nonnegative matrix factorization with the β-divergence

Arthur Marmin, José Henrique de Morais Goulart, Cédric Févotte

Abstract: Highlights•Most nonnegative matrix factorization algorithms are based on block coordinate descent.•Updating the factors jointly can save up computational time.•Joint majorization-minimization results in new and simple multiplicative updates.•Computational savings are significant for the quadratic, Kullback-Leibler and Itakura-Saito losses.