Can any node classification model consider edge features?


The current popular GNN models, like GAT, only focus on node features and largely ignore the edge features that contain rich information.

Is there any tutorial or public code about this kind of node-classification model considering both node features and edge features during calculation?

Looking forward to your comments.

Best wishes,
Zijian Wang

Have you checked FAQ ? Point 11 is related to your issue.

1 Like

Hi Zijian,

If your edge features are just edge weights, you can weight the messages as @flpgrz suggested. If your edge features are not edge weights, you may check point 13 in FAQ.

1 Like

There are some hacks that give you a way to do this. Turn your edges into nodes.

1 Like

Hi flpgrz,

Thanks for your reply. I had a check, the FAQ is really useful.

Hi Mufei Li,

Thanks so much for your reply. The answers in FAQ are really practical to me.
Sorry I did not read the FAQ at the beginning, where all my desired answer is in it.

Have a nice day/night.

1 Like

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.