Saul Imbrici

Graph Neural Networks for Scalable and Transferable Node Centrality Approximation

Results show that mixed-distribution training can improve structural transfer in GNN-based centrality approximation, while identifying closeness centrality's sensitivity to topology as an open challenge.