Multidimensional edge features

benedekrozemberczki / pytorch_geometric_temporal

PyTorch Geometric Temporal: Spatiotemporal Signal Processing with Neural Machine Learning Models (CIKM 2021)

MIT License

2.63k stars 370 forks source link

Multidimensional edge features #232

Closed Alessiobrini closed 1 year ago

Alessiobrini commented 1 year ago

Is it possible that none of the models implemented accepts multidimensional edge features or more general edge features? I see just the edge weights argument in each forward method implemented. Can someone help me clarify this?

SimonPop commented 1 year ago

Hi!

I think you are right, at the moment there does not seem to be any model taking edge features as a possible input to their forward method.

Actually, the only layers currently in use from PyG I could find are:

GCNConv
ChebConv
SageConv although with HeteroConv that allows for edge attributes.
GatedGraphConv
RGCNConv

None of these allow for edge features, hence your conclusion. Although, some allow for edge weights (e.g. GatedGraphConv), and others for edge type (e.g. RGCNConv)