Review for NeurIPS paper: Adversarially-learned Inference via an Ensemble of Discrete Undirected Graphical Models