Reviews: Spectral Learning of Dynamic Systems from Nonequilibrium Data

Update after author feedback: I thank the authors for their constructive response and hope they can incorporate as many of the promised changes as possible. Based on the feedback and reviewer discussion I now have a better idea of the motivation behind the work. Perhaps a brief description of a concrete example application problem motivating the work would make it easier for others more used to machine learning type learning of dynamical models to appreciate the work too. The experimental evaluation would benefit a lot from explicit comparisons with Bayesian alternatives (e.g. Ruttor et al., NIPS 2013; Svensson et al., AISTATS 2016 and references therein) to properly understand the pros and cons of the different approaches.