search

divelab / DIG

A library for graph deep learning research
https://diveintographs.readthedocs.io/
GNU General Public License v3.0
1.84k stars 281 forks source link

Using PYG edge_weight instead of defining custom propagate #99

Closed steveazzolin closed 2 years ago

steveazzolin commented 2 years ago

Hi all,

I have the following question:

I saw that for PGExplainer you defined a custom propagate function in order to account for the edge matrix computed by the explainer, and this is fine. But, is it possible to achieve the same result by leaving the default propagate implementation and feeding the weights to the network with the additional parameter edge_weight? Specifically, I am referring for simplicity just to the GCN architecture.

Thanks!

Oceanusity commented 2 years ago

Thank you for your issue. For GCN, it can use the edge_weight to achieve similar propagation, and the implementation is here.