Characterizing the Effects of Model Parameters on Performance of Convolutional Neural Networks for Cosmic Ray Shower Reconstruction

Abstract

In collaboration with IceCube South Pole Neutrino Observatory, data from over 400,000 cosmic ray energy shower events recorded by IceTop, the surface component of IceCube, were used to train an array of convolutional neural network (CNN) models that reconstruct the initial energy of the cosmic ray primary.