Generative Adversarial Networks and Markov Random Fields for oversampling very small training sets | Publicación