Can one compose lux layers with graph neural network

[jakubMitura14]

Can one compose lux layers with graph neural network.

For flux there is geometric flux, but composing lux and flux is sth to be avoided

Or alternatively can one build a model in lux and incorporate it as a part of flux chain , to compose it with flux geometric

[jakubMitura14]

I need to experiment more first

[YichengDWu]

GeometricFlux is heavily flawed and I wouldn't recommend anyone to use it. You can check how my package NeuralGraphPDE defines a graph neural network, it's very simple

[jakubMitura14]

Fantastic thank you !