NeurIPS 2020

Self-supervised Auxiliary Learning with Meta-paths for Heterogeneous Graphs

Meta Review

Overall, this paper was borderline. The main strengths of the paper is that it's addressing an important problem, and that the experiments and presentation of the work are both solid. The reviewers appreciated the new experiments in the author response. The main weaknesses of the paper are the novelty of the overall idea and the increased efforts required for customized model design for different datasets/graphs. In the balance, I think that the merits outweigh the flaws.