revised version of our article. 2 R# 2: My only relatively minor issue with the paper range of the earlier layers)

W e thank all reviewers for their comments. As you point out, we incorrectly asserted that it suffices to consider a one-layer neural network. Thank you for this interesting comment, we will clarify/fix these issues in the final version of our paper. I think that in Line 84, the authors should add a reference to "A Geometric Analysis of Phase Retrieval". R#3: We really appreciate your positive feedback!