Testing of KNO-scaling of charged hadron multiplicities within a Machine Learning based approach

TL;DR

This paper employs deep neural networks to analyze KNO-scaling in charged hadron multiplicities across different LHC energies, demonstrating that the scaling is preserved at the hadronic level despite variations in jet multiplicity scaling.

Contribution

It introduces a machine learning approach using deep residual networks to study KNO-scaling in high-energy particle collisions, a novel application in this context.

Findings

KNO-scaling is preserved at the hadronic level.

Machine learning models can predict multiplicity distributions across energies.

Scaling of mean jet multiplicity varies with model complexity.

Abstract

The results of a Machine Learning-based method is presented here to investigate the scaling properties of the final state charged hadron and mean jet multiplicity distributions. Deep residual neural network architectures with different complexities are utilized to predict the final state multiplicity distribution from the parton-level final state, generated by the \textsc{Pythia} Monte Carlo event generator. Hadronization networks were trained by $\sqrt{s}=7$ TeV events, while predictions have been made for various LHC energies from $\sqrt{s}=0.9$ TeV to 13 TeV. Scaling properties were adopted by the networks at hadronic level, indeed KNO-scaling is preserved -- although, the scaling of the mean jet multiplicity distributions varies for the applied models.