python3, tensorflow==1.10.0
GNN: is that you use a RNN to memory every possible step of your input node, the target node is the answer.
GCN: you can see the pic
folder.
task/model | 1-step-gnn | 2-step-gnn |
---|---|---|
babi-4 | 100% | 100% |
babi-15 | 0% | 100% |
GATED GRAPH SEQUENCE NEURAL NETWORKS
Semi-Supervised Classification with Graph Convolutional Networks