Congratulations! This paper was a borderline case that led to significant discussion among the reviewers, but in the end we decided the novel ODE-style approach would be of strong interest to the NeurIPS community. In your revision, please address all promises in the rebuttal and comments in the reviews. If there are additional experiments or examples you can include to demonstrate practical value or application of this technique to machine learning tasks, please include these in the revision to better demonstrate where your method might be used---this was the main weakness identified by the reviewers.