GNN regression and DAG, how to constrain which features directly influence the target values and which influence indirectly

[bernddude]

Hello, i am trying to use a GAT or GCN to perform a regression
I have a DAG which describes the influence of the features among each other and on the regression target

I understand how i can us the edge list to describe the interaction of the features
What i am a bit at a loss is how i express the interaction between the regression target and the features ?

for example say i have three features A, B and C and a target Y
my DAG says that A -> C and B ->C and C -Y and B-> Y directly

how would i set this up using the existing pytorch_geometric API ?

thank you so much for any advise , Bernd

[rusty1s]

In this case, you would need to represent your target Y as a readout node inside your graph, so A, B, C and Y are all nodes in your network.