NeurIPS 2020

Towards Deeper Graph Neural Networks with Differentiable Group Normalization

Meta Review

This paper presents a method to address the over-smoothing issue in deep graph neural network with differentiable group normalization and metrics to measure over-smoothing. All reviewers agree that this paper tackles an important problem and the empirical results verify the main claim of the paper. The reviewers raised some issues regarding comparisons with previous work. However, the authors noted that these papers are relatively recent (available publicly in April 2020 or later). I consider these papers to be parallel work and therefore understand why the authors did not compare with them in the current version. I encourage the authors to include these comparisons for completeness in the final version, as well as other feedbacks around analysis from R2 and R5.