Hello,
when implementing the model of the paper https://arxiv.org/pdf/1811.05320.pdf why do you use a GCNConv for each update and reset gate and candidate state? Why not just take the output of a single GCNConv as input for the GRU part (as they do in the paper)?
Hello, when implementing the model of the paper https://arxiv.org/pdf/1811.05320.pdf why do you use a
GCNConv
for eachupdate
andreset
gate andcandidate
state? Why not just take the output of a singleGCNConv
as input for theGRU
part (as they do in the paper)?