196f5641aa9dc87067da4ff90fd81e7b-AuthorFeedback.pdf

AC and all Reviewers: We thank all reviewers. We discuss the motivation of using S-based prior later. In fact, using any prior at all (i.e., non-uniform) is optional in our formulation, Our method is used for inference only, and can work on top of any pre-trained feature extractor. Is the method limited to uniform distribution? No, we can use any prior in Eq. (2).