geometric-machine-learning

This paper introduces Equivariant Sheaf Neural Networks (ESNN), which learn directed geometric transport on graphs to enhance equivariant message passing for physical systems while preserving exact Euclidean symmetry.