be53d253d6bc3258a8160556dda3e9b2-AuthorFeedback.pdf

We thank all the reviewers for acknowledging the novelty and concreteness of the path-integral idea presented in1 thispaper. Wefinditveryencouraging thatthereviewersfound thepresentation clearly connects ourgraphlearning2 framework with its deep roots in physics, and that the experimental results are convincing. We indeed believe in3 the conceptual and practical values of the PAN framework. We hope the new data set can also contribute to the4 community. In fact, as Reviewer #2 pointed out (also mentioned in related works), the advantage of10 the PAN convolution idea in node classification tasks have been reported in a short ICML workshop paper, see11 arXiv:1904.10996.