On Provable Benefits of Depth in Training Graph Convolutional Networks

Open in new window