Using a variable graph structure with NeuralODE and GCNConv

Specific Domains Machine Learning
, , ,
2

Hi! I am glad to see feedback from you.
It’s flexible to adapt variable graph strategy to several graph convolution layers.
Unfortunately, currently variable graph strategy is not supported for GNODE.
If you’re interested in this approach, it’s welcome to fire an issue to GeometricFlux.jl.

1 Like