Improving Classification of Ultra-High Energy Cosmic Rays Using Spacial Locality by Means of a Convolutional DNN

Abstract

Machine learning algorithms have shown their usefulness in a countless variety of fields. Specifically in the astrophysics field, these algorithms have helped the acceleration of our understanding of the Universe and the interaction between particles in recent years. Deep learning algorithms, enclosed in machine learning field, are showing outstanding performance in problems where spatial information is crucial, such as images or data with time dependency. Cosmic rays are high-energy radiation, mainly originated outside the Solar System and even from distant galaxies that constitute a fascinating problem in Physics today. When a Ultra-High Energy Cosmic Ray enters the Earth’s atmosphere an extensive air shower is generated. An air shower is a cascade of particles and can be recorded with surface detectors. This work develops a supervised learning algorithm to classify the signals recorded by surface detectors with the aim of identifying the primary particle giving rise to the extensive air shower. Convolutional Neural Networks along with Feed Forward Neural Networks will be compared. Also, the aggregation of information from different surface detectors recording the same phenomenon will be studied against using the information of a single surface detector.