Review for NeurIPS paper: Efficient active learning of sparse halfspaces with arbitrary bounded noise