Hi, I'm wondering which function did you use to generate the visualization on MUTAG? (Figure 4 and 5 in the paper)
Or if authors have any follow-up on this?
Some other detailed questions:
You are using the dense version of GCN, where adj can be float and it won't affect matmul operation (like this line). But if we are moving to the sparse version of GCN (like pyg), then how are you going to handle the grad according the edge_list which is essentially the node index (like this line)?
Hi, I'm wondering which function did you use to generate the visualization on MUTAG? (Figure 4 and 5 in the paper) Or if authors have any follow-up on this?
Some other detailed questions:
adjcan befloatand it won't affectmatmuloperation (like this line). But if we are moving to the sparse version of GCN (like pyg), then how are you going to handle thegradaccording theedge_listwhich is essentially the node index (like this line)?