The evolved W that comes from the GRU unit has the same shape as X which is (nodes, features). In the GConv layer, we are passing W and performing a multiplication X@W but we should be performing X@W.T for the shapes to match. It happens that here W is square (due to TopKPooling) which is why it might have been missed. Therefore, I believe we need to pass W.squeeze(dim=0).T. Same goes for the O-version of EvolveGCN.
I believe there is a bug in the following line.
https://github.com/benedekrozemberczki/pytorch_geometric_temporal/blob/e43e95d726174574eeae32ad8c4670631fd0a58d/torch_geometric_temporal/nn/recurrent/evolvegcnh.py#L98
The evolved
Wthat comes from the GRU unit has the same shape asXwhich is (nodes, features). In the GConv layer, we are passingWand performing a multiplicationX@Wbut we should be performingX@W.Tfor the shapes to match. It happens that hereWis square (due to TopKPooling) which is why it might have been missed. Therefore, I believe we need to passW.squeeze(dim=0).T. Same goes for the O-version of EvolveGCN.Could someone verify this please.
Thank you.