arxivhep-thcs.LG2026-07-03
Graph Neural Networks for the Graphical Bootstrap
Rigers Aliaj, Gabriele Dian, Reza Doobary, Paul Heslop
We study a graph classification problem involving over 20 million graphs, arising from high-order perturbative computations of correlators in planar $\mathcal{N}=4$ super-Yang--Mills, a model closely related to the theory of the strong nuclear force. We benchmark graph neural net…