Normalizing flows for lattice quantum field theory

Lattice quantum field theory (QFT) is an integral part of the modern nuclear and particle physics toolkit. It is a computational framework based around Monte Carlo sampling in billions of dimensions to approximate functional "path integrals", requiring large-scale supercomputing in practice. Accelerating or otherwise improving this sampling task provides a challenging testing ground for generative AI/ML methods, with unusual requirements in model structure and statistical exactness. I discuss ongoing efforts to use normalizing flows and their generalizations to improve calculations in lattice quantum chromodynamics (QCD), the theory of the strong force.