Evaluating EAS Directions from TAIGA HiSCORE Data Using Fully Connected Neural Networks

Abstract

The TAIGA-HiSCORE setup is a wide-angle Cherenkov detector array for recording extensive air showers (EASs). The array comprises over 120 stations located in the Tunka Valley near Lake Baikal. One of the main tasks of data analysis in the TAIGA-HiSCORE experiment is to determine the axis direction of the EASs and their core location. These parameters are used to determine the source of gamma rays and play an important role in estimating the energy of the primary particle. The data collected by HiSCORE stations include signal amplitude and arrival time and allow for estimation of the shower direction of arrival. In this work, we use Monte Carlo simulation data for HiSCORE to demonstrate the feasibility of determining the EAS axis directions with artificial neural networks. Our approach employs multilayer perceptrons with skip connections, which take data from subsets of HiSCORE stations as input. Multiple station subsets are selected to derive more accurate composite estimates. Furthermore, we use a two-stage algorithm, where the initial direction estimates in the first stage are refined in the second stage. The final estimates have an average error of less than 0.25\({}^{\circ}\). We plan to use this approach as a part of multimodal analysis of data obtained from several types of detectors used in the TAIGA experiment.