Particle identification in the GlueX detector with machine learning

TL;DR

This paper demonstrates that machine learning algorithms like multilayer perceptrons and boosted decision trees significantly improve particle identification accuracy in the GlueX experiment compared to traditional cut-based methods.

Contribution

It introduces machine learning techniques to enhance particle identification in the GlueX detector, surpassing existing cut-based approaches.

Findings

ML algorithms outperform traditional cut-based PID methods

Both perceptrons and decision trees achieve higher accuracy in simulated data

Significant improvement in identifying charged and neutral particles

Abstract

In particle physics experiments, identifying the types of particles registered in a detector is essential for the accurate reconstruction of particle collisions. At Thomas Jefferson National Accelerator Facility (Jefferson Lab), the GlueX experiment performs particle identification (PID) by setting specific thresholds, known as cuts, on the kinematic properties of tracks and showers obtained from detector hits. Our research aims to enhance this cut-based method by employing machine-learning algorithms based on multi-layer perceptrons and boosted decision trees. Similar approaches have been applied in other particle physics experiments and offer an opportunity to increase PID accuracies using reconstructed kinematic data. Our study illustrates that both multilayered perceptrons and boosted decision trees can identify charged and neutral particles in Monte Carlo simulated GlueX data with significantly improved accuracy over the current cuts-based PID method.