I’m trying to do a homogeneous graph, and now I have both node features and edge features, how can I use them both to accomplish this task? Actually, I’m going to use the EGATConv provided by DGL to reach my goal, maybe I should Readout the node features and edge features separately, then concatenate, and finally softmax classification?
ndata and edata are always available during message passing. As for how to use them, it mainly depend on your task and model you’d like to apply. DGL provides example for EGATConv here.