Tackling the muon identification in water Cherenkov detectors problem for the future Southern Wide-field Gamma-ray Observatory by means of Machine Learning

TL;DR

This paper explores machine learning techniques to improve muon identification in water Cherenkov detectors with reduced water volume, enabling cost-effective gamma-ray observatories and advancing cosmic ray research.

Contribution

It introduces new feature engineering and ML approaches tailored for low-volume Cherenkov detectors, demonstrating effective muon identification and potential for cost savings.

Findings

Machine learning achieves high muon identification accuracy

New features improve model generalization

Reduced water volume detectors can effectively discriminate muons

Abstract

This paper presents several approaches to deal with the problem of identifying muons in a water Cherenkov detector with a reduced water volume and 4 PMTs. Different perspectives of information representation are used and new features are engineered using the specific domain knowledge. As results show, these new features, in combination with the convolutional layers, are able to achieve a good performance avoiding overfitting and being able to generalise properly for the test set. The results also prove that the combination of state-of-the-art Machine Learning analysis techniques and water Cherenkov detectors with low water depth can be used to efficiently identify muons, which may lead to huge investment savings due to the reduction of the amount of water needed at high altitudes. This achievement can be used in further research to be able to discriminate between gamma and hadron induced showers using muons as discriminant.