Adversarially-Trained Nonnegative Matrix Factorization

We consider an adversarially-trained version of the nonnegative matrix factorization, a popular latent dimensionality reduction technique. In our formulation, attacker adds arbitrary bounded norm to given data matrix. design efficient algorithms inspired by adversarial training optimize for dictionary and coefficient matrices with enhanced generalization abilities. Extensive simulations on synthetic benchmark datasets demonstrate superior predictive performance completion tasks proposed method compared state-of-the-art competitors, including other variants factorization.