Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

GINEConv edge_attributes and node_features

See original GitHub issue

❓ Questions & Help

Greetings,

I’m trying to use GINEConv for graph classification, but it seems to require the same number of node_features and edge_attributes in order to work correctly. When this number is not the same, I get the following error assert x.size(-1) == edge_attr.size(-1) am I using it wrong or is it indeed a requirement to have the same number of node and edge features?

Thank you for your help.

Issue Analytics

State:
Created 3 years ago
Comments:7 (3 by maintainers)

Top GitHub Comments

1reaction

rusty1scommented, May 3, 2021

Yes, this needs to be added to every layer.

1reaction

rusty1scommented, Jun 19, 2020

GIN-E requires node and edge features to be of same dimensionality, since their features are added together. In general, this isn’t much of a problem since you can always transform your edge features to the same dimensionality.

Top Results From Across the Web

torch_geometric.nn — pytorch_geometric documentation

Constructs messages from node \(j\) to node \(i\) in analogy to \(\phi_{\mathbf{\Theta}}\) for each edge in edge_index . This function can take any...

GINEConv edge_attributes and node_features #1342 - GitHub

Questions & Help Greetings, I'm trying to use GINEConv for graph classification, ... GIN-E requires node and edge features to be of same ......

How to use edge features in Graph Neural Networks (and ...

In this video I talk about edge weights, edge types and edge features and how to include them in Graph Neural Networks.

GINEConv — DGL 0.9.1post1 documentation

Graph Isomorphism Network with Edge Features, introduced by Strategies for Pre-training ... If not None, it will be applied to the updated node...

Incorporate Node and Edge Features in Graph Neural Networks

network, named NENN, which incorporates node and edge features based on a dual-level attention mechanism, including node-level and ... molecular properties.