search

CarloLucibello / GraphNeuralNetworks.jl

Graph Neural Networks in Julia
https://carlolucibello.github.io/GraphNeuralNetworks.jl/dev/
MIT License
214 stars 47 forks source link

Convolutions for GNNHeteroGraphs #254

Closed CarloLucibello closed 1 year ago

CarloLucibello commented 1 year ago

Now we have GNNHeteroGraphs but none of the convolutional layers support them.

See https://pytorch-geometric.readthedocs.io/en/latest/notes/heterogeneous.html and in particular to_hetero for the interface PyG provides.

Dsantra92 commented 1 year ago

I am willing to do this.

arnauqb commented 1 year ago

Has there been any progress on this? Am I correct on assuming that Message Passing is not yet supported for heterogeneous graphs? Thanks!

Dsantra92 commented 1 year ago

Yes. Only the basic structure of HeteroGNNGraphs is complete till date. Various graph functions and convolutions are not yet supported.

CarloLucibello commented 1 year ago

a preliminary implementation of a convolution for heterographs has been added in #300. Many rough edges though, feedbacks welcome.