Speaker
Description
I will present a neural-network–based framework for modeling generalized parton distributions, referred to as NNGPD, in which GPDs are represented as flexible functions constrained through physically motivated integral relations. This approach reflects the inverse-problem character of GPD phenomenology - without assuming a specific functional ansatz - where experimental and theoretical information is incorporated into the training procedure via loss functions enforcing convolution integrals (Compton form factors), as well as Mellin moments accessible in lattice QCD. We find that the neural-network representation reproduces the main features of the GPDs over the relevant kinematic domain, despite being constrained only by their integral projections thus demonstrating its viability for constraining GPDs through global physical observables. A thorough analysis of both epistemic and statistical uncertainty, the latter being inherent to the sampling of allowed GPD solutions, will be discussed. Our approach provides a basis for future phenomenological applications incorporating experimental measurements, including those anticipated at the Electron–Ion Collider.
| Speaker confirmation | Yes |
|---|