Applications of AI/ML in lattice QCD calculations of hadron physics

Understanding nucleons, the building blocks of the visible universe, including their spin and mass structures governed by gluons—the mediator bosons of the strong interaction—remains one of the grand challenges in nuclear and particle physics. Additionally, unraveling the processes behind the matter-antimatter asymmetry in the universe is of fundamental importance, with elusive neutrinos potentially holding the key to this mystery. These research areas are central to the physics objectives of the future Electron-Ion Collider (EIC) and the Deep Underground Neutrino Experiment (DUNE). While Lattice QCD calculations of hadron structures have made remarkable progress in determining parton distribution functions (PDFs), generalized parton distributions (GPDs), and recently nucleon transition matrix elements relevant to neutrino physics, these calculations face significant challenges. These include extracting continuous distributions from discrete and limited data points, restricted access to large hadron momenta, and addressing inverse problems along with uncertainty quantification. To address these issues, we explore the synergy between Lattice QCD calculations of hadron structures and generative machine learning algorithms. We highlight examples of how this synergy can enhance the determination of both polarized and unpolarized gluon PDFs, as well as its potential application for extracting GPDs from lattice data.