Review for NeurIPS paper: Characterizing emergent representations in a space of candidate learning rules for deep networks