Crate propago 

propago: The Learning Layer.

Provides a small set of graph learning primitives built on candle tensors.

Current public surface:

• GCNConv: a simple graph convolution (linear + adjacency matmul)
• HGCNConv: hyperbolic graph convolution on the Poincaré ball (Tensor-native ops)

Re-exports

pub use hyperbolic::HGCNConv;

pub use nn::GCNConv;

Modules

hyperbolic

nn