1 paper
Patrick Gillespie, Layal Bou Hamdan, Ioannis Schizas +2
Equipping graph neural networks with a convolution operation defined in terms of a cellular sheaf offers advantages for learning expressive representations of heterophilic graph da…