Input of square of adjacency matrix in GCN

[shrimonmuke0202]

How can I feed the square of the adjacency matrix of a graph in the GCN module?

[rusty1s]

You can directly pre-compute that, e.g., by utilizing torch-sparse:

edge_index, _ = torch_sparse.spspmm(edge_index, None, edge_index, None, N, N, N)

where N denotes the number of nodes.

[shrimonmuke0202]

Can I also pass the cube of the adjacency matrix using same way ?

…

On Thu, 22 Apr 2021, 12:52 pm Matthias Fey, ***@***.***> wrote: You can directly pre-compute that, e.g., by utilizing torch-sparse: edge_index, _ = torch_sparse.spspmm(edge_index, None, edge_index, None, N, N, N) where N denotes the number of nodes. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#2442 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AQ7MTCWQTS5LN7EVTZLYZX3TJ7FDRANCNFSM43LBYJUA> .

[rusty1s]

Yes, but keep in mind that your adjacency matrix can get pretty dense :)

[shrimonmuke0202]

Can you give me an example ?

…

On Thu, 22 Apr 2021, 1:03 pm Matthias Fey, ***@***.***> wrote: Yes, but keep in mind that your adjacency matrix can get pretty dense :) — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#2442 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AQ7MTCW6PAVUVQVAQKDK2MLTJ7GMPANCNFSM43LBYJUA> .