Review for NeurIPS paper: Efficient Variational Inference for Sparse Deep Learning with Theoretical Guarantee