Deep learning for exploring hadron-hadron interactions

TL;DR

This paper explores the application of deep learning techniques to study hadron-hadron interactions, utilizing supervised and unsupervised neural networks to extract interaction potentials from experimental and simulation data, advancing understanding in particle physics.

Contribution

It introduces novel deep learning approaches for deriving hadron interaction potentials from collision experiments and lattice QCD simulations, bridging experimental data and theoretical models.

Findings

Deep neural networks successfully learn inverse mappings from collision data to interaction potentials.

Unsupervised models effectively construct potential functions directly from simulated correlation functions.

Deep learning methods demonstrate significant potential in advancing hadron interaction studies.

Abstract

In this proceeding, we introduce deep learning technologies for studying hadron-hadron interactions. To extract parameterized hadron interaction potentials from collision experiments, we employ a supervised learning approach using Femtoscopy data. The deep neural networks (DNNs) are trained to learn the inverse mapping from observations to potentials. To link between experiments and first-principles simulations, we further investigate hadronic interactions in Lattice QCD simulations from the HAL QCD method perspective. Using an unsupervised learning approach, we construct a model-free potential function with symmetric DNNs, aiming to learn hadron interactions directly from simulated correlation functions (equal-time Nambu-Bethe-Salpeter amplitudes). On both fronts, deep learning methods show great promise in advancing our understanding of hadron interactions.