NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:3549
Title:Diffeomorphic Temporal Alignment Nets

This paper developed a deep learning approach to aligning time series by incorporating a diffeomorphism. The reviewers found the paper enjoyable to read as the method was clearly explained and their were nice visualizations to present the intuition. The majority of the reviewers thought that the experiments were thorough enough to demonstrate the efficacy of the algorithm, however, one reviewer would have liked to see the method used for something beyond time-series classification. The author response did not really address this point to the reviewer's satisfaction, so the authors should consider this for the camera-ready version of the paper. Finally, the reviewers pointed out that the paper has some notational issues and the intro need some work to motivate the work and provide background. The authors should take these comments into account for the camera-read version to improve the quality of the paper.