A Simple Spectral Failure Mode for Graph Convolutional Networks

10/25/2020

∙

by   Carey E. Priebe, et al.

∙

2

∙

share

We present a simple generative model in which spectral graph embedding for subsequent inference succeeds whereas unsupervised graph convolutional networks (GCN) fail. The geometrical insight is that the GCN is unable to look beyond the first non-informative spectral dimension.