Review for NeurIPS paper: Efficient Learning of Discrete Graphical Models