NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:6006
Title:Layer-Dependent Importance Sampling for Training Deep and Large Graph Convolutional Networks


		
All reviewers agreed that this work addresses a highly relevant topic, and that the theoretical results regarding the variance reduction is certainly interesting. There was, on the other hand, some disagreement about the "true" novelty and the significance of the differences with respect to related approaches, which could not be fully resolved in the discussions following the author rebuttal -- but such a controversial discussion might also be seen as an indicator for the overall importance of this paper.