Parametrizing Compton form factors with neural networks

TL;DR

This paper introduces a neural network-based method to extract Compton form factors in the deeply virtual region, offering an alternative to traditional model fitting techniques, demonstrated on toy models and HERMES data.

Contribution

The paper presents a novel neural network approach for parametrizing Compton form factors, improving flexibility over standard fitting methods.

Findings

Neural network method effectively recovers form factors from simulated data.

Approach successfully applied to real HERMES experimental data.

Compared favorably to traditional least-squares fitting.

Abstract

We describe a method, based on neural networks, of revealing Compton form factors in the deeply virtual region. We compare this approach to standard least-squares model fitting both for a simplified toy case and for HERMES data.