Graph neural network

A graph neural network (GNN) is a type of neural network which operates on a graph structure. The model captures dependencies of graphs via message passing between the nodes of graph. Use cases include modeling physical systems including physics and biological models.