by Mila / University of Cambridge · Graph and Network Models
A neural network architecture using masked self-attention on graph-structured data, learning different weights for different neighbor nodes.
No problems reported for this model.