Abstract

Graph learning is currently dominated by graph kernels, which, whilepowerful, suffer some significant limitations. Convolutional Neural Networks(CNNs) offer a very appealing alternative, but processing graphs with CNNs isnot trivial. To address this challenge, many sophisticated extensions of CNNshave recently been introduced. In this paper, we reverse the problem: ratherthan proposing yet another graph CNN model, we introduce a novel way torepresent graphs as multi-channel image-like structures that allows them to behandled by vanilla 2D CNNs. Experiments reveal that our method is more accuratethan state-of-the-art graph kernels and graph CNNs on 4 out of 6 real-worlddatasets (with and without continuous node attributes), and close elsewhere.Our approach is also preferable to graph kernels in terms of time complexity.Code and data are publicly available.

 

Introduction (beta)

None

 

Conclusion (beta)

None