New Particle Identification Approach with Convolutional Neural Networks in GAPS

TL;DR

This paper introduces a novel 3D convolutional neural network approach for particle identification in the GAPS experiment, enhancing accuracy over previous machine learning methods by leveraging energy deposition data.

Contribution

The paper presents a new 3D CNN model that analyzes energy depositions for antiparticle identification, improving performance in the GAPS experiment.

Findings

The CNN outperforms existing machine learning methods.

Combining physical quantities with CNN improves identification accuracy.

The approach effectively captures positional and energy correlations.

Abstract

The General Antiparticle Spectrometer (GAPS) is a balloon-borne experiment that aims to measure low-energy cosmic-ray antiparticles. GAPS has developed a new antiparticle identification technique based on exotic atom formation caused by incident particles, which is achieved by ten layers of Si(Li) detector tracker in GAPS. The conventional analysis uses the physical quantities of the reconstructed incident and secondary particles. In parallel with this, we have developed a complementary approach based on deep neural networks. This paper presents a new convolutional neural network (CNN) technique. A three-dimensional CNN takes energy depositions as three-dimensional inputs and learns to identify their positional/energy correlations. The combination of the physical quantities and the CNN technique is also investigated. The findings show that the new technique outperforms existing machine learning-based methods in particle identification.