Understanding the Representation Power of Graph Neural Networks in Learning Graph Topology

We find that GCNs are rather restrictive in learning graph moments. Without careful design, GCNs can fail miserably even with multiple layers and nonlinear activation functions.