Speaker
Description
We discuss the use of machine learning techniques for the modeling of generalized parton distributions in view of their nonparametric estimation from experimental data. Current GPD extractions indeed suffer from a model dependence which lessens their impact and brings unknown systematics in the estimation of derived quantities like 3D tomography or angular momentum decomposition. On the contrary this new strategy to describe GPDs allows a flexible implementation of theory driven constraints and provides tools to keep model dependence at a minimum level. We also address aspects of a practical nature like the design and training of artificial neural networks suitable for this analysis. Getting a better grip on the control of systematic effects, our work will help GPD phenomenology to achieve its maturity in the precision era of GPD extractions opened by a new generation of experiments.
