Physics-Informed Neural Networks for the nucleon structure

The structure of the nucleon is encoded in functions that are usually extracted from the experimental data and obey certain integro-differential equations. Solving the inverse problem is a non-trivial task, and methods of machine learning promise to facilitate it. In our study, we apply Physics-Informed Neural Networks (PINN) to the solution of integro-differential equations, characteristic of parton densities of the nucleon. We start from an example of solutions of Maxwell's equations and then apply PINNs to solutions of DGLAP equations that encode the scale dependence of collinear parton densities.