Extraction of Information from Polarized Deep Exclusive Scattering with Machine Learning (2406.09258v1)

Abstract: A framework defining benchmarks for the analysis of polarized exclusive scattering cross sections is proposed that uses physics symmetry constraints as well as lattice QCD predictions. These constraints are built into ML algorithms. Both physics driven and ML based benchmarks are applied to a wide range of deeply virtual exclusive processes through explainable ML techniques with controllable uncertainties. The observables, namely the Compton Form Factors (CFFs) which are convolutions of Generalized Parton Distributions (GPDs), are extracted using methods such as the random targets method to evaluate the separate contribution of the aleatoric and epistemic uncertainties in exclusive scattering analyses.