Deep Learning for Parton Distribution Extraction
Analysis
This paper introduces a novel machine-learning method using neural networks to extract Generalized Parton Distributions (GPDs) from experimental data. The method addresses the challenging inverse problem of relating Compton Form Factors (CFFs) to GPDs, incorporating physical constraints like the QCD kernel and endpoint suppression. The approach allows for a probabilistic extraction of GPDs, providing a more complete understanding of hadronic structure. This is significant because it offers a model-independent and scalable strategy for analyzing experimental data from Deeply Virtual Compton Scattering (DVCS) and related processes, potentially leading to a better understanding of the internal structure of hadrons.
Key Takeaways
- •Presents a machine-learning method for extracting GPDs from experimental data.
- •Uses a neural network with a physics-preserving layer for the QCD kernel.
- •Provides a probabilistic extraction of GPDs.
- •Offers a model-independent and scalable strategy for analyzing DVCS data.
“The method constructs a differentiable representation of the Quantum Chromodynamics (QCD) PV kernel and embeds it as a fixed, physics-preserving layer inside a neural network.”