Review for NeurIPS paper: Discovering Symbolic Models from Deep Learning with Inductive Biases