From quarks and gluons to hadrons - Jure Zupan (U. Cincinnati)

Abstract: Monte Carlo event generators for particle collisions are composed of three block, the calculations of hard matrix elements, parton shower, and hadronization. While the first two are theoretically under good control and systematically improvable using perturbative techniques, hadronization relies on the use of phenomenological models. I will review the first attempts to use Machine Learning architectures to describe hadronization, with the ultimate goal to train directly on data. The first practical side product of this effort is an algorithm for  faster evaluation of uncertainties associated with the Lund string model implemented in Pythia.