robust multi-object matching
Review for NeurIPS paper: Robust Multi-Object Matching via Iterative Reweighting of the Graph Connection Laplacian
Weaknesses: - The presentation can be largely improved: some of the sentences do not help the reader and currently the draft is not easy to read. For instance, the transition to the second paragraph of the introduction is quite abrupt: while the first paragraph talks about keypoints and images, the keypoints disappear when talking about the permutation synchronization. Moreover, the sigma is not properly defined. Similarly, Section 4.1 is hard to read: while the overall idea is clear, the reader gets trapped in too many details, such as those in lines 196-199. Similar comments hold for Section 4.2: why presenting two techniques when you recommend using only one? While I understand the desire of the authors to be comprehensive, mentioning too many details may compromise clarity.