Graph Neural Networks: A learning journey since 2008
Secondly, we can define the layer-wise propagation rule in GCN, which is eq.10. This function takes as input the adjacency matrix, the degree matrix, the graph features matrix and the i-th layer weights. Finally, initialise layers weights and set up 2 layers — but you can set up as many layers as you want — and give GCN a try!
DA: 68 PA: 79 MOZ Rank: 34