Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Evaluation of EAS directions based on TAIGA HiSCORE data using fully connected neural networks

A. P. Kryukov, S. P. Polyakov, Yu. Yu. Dubenskaya, E. O. Gres, E. B. Postnikov, P. A. Volchugov, D. P. Zhurov

TL;DR

This paper tackles the problem of precisely reconstructing extensive air shower directions from TAIGA HiSCORE detector data to aid gamma-ray source localization. It introduces a two-stage, fully connected neural network framework with skip connections that operates on subsets of triggered stations, where the first stage yields preliminary directions and the second stage refines them after transforming inputs onto a plane orthogonal to the estimated direction. The method achieves sub-degree angular accuracy, with a mean error around 0.215 degrees on the primary test set and as low as 0.0885 degrees for events with many triggered stations, demonstrating performance comparable to conventional approaches and enabling multimodal analyses across TAIGA detectors. The work highlights the potential of subset-based, two-stage neural pipelines for high-precision EAS direction reconstruction in large detector arrays and sets the stage for integrating data from multiple detector types in TAIGA.

Abstract

The direction of extensive air showers can be used to determine the source of gamma quanta and plays an important role in estimating the energy of the primary particle. The data from an array of non-imaging Cherenkov detector stations HiSCORE in the TAIGA experiment registering the number of photoelectrons and detection time can be used to estimate the shower direction with high accuracy. In this work, we use artificial neural networks trained on Monte Carlo-simulated TAIGA HiSCORE data for gamma quanta to obtain shower direction estimates. The neural networks are multilayer perceptrons with skip connections using partial data from several HiSCORE stations as inputs; composite estimates are derived from multiple individual estimates by the neural networks. We apply a two-stage algorithm in which the direction estimates obtained in the first stage are used to transform the input data and refine the estimates. The mean error of the final estimates is less than 0.25 degrees. The approach will be used for multimodal analysis of the data from several types of detectors used in the TAIGA experiment.

Evaluation of EAS directions based on TAIGA HiSCORE data using fully connected neural networks

TL;DR

This paper tackles the problem of precisely reconstructing extensive air shower directions from TAIGA HiSCORE detector data to aid gamma-ray source localization. It introduces a two-stage, fully connected neural network framework with skip connections that operates on subsets of triggered stations, where the first stage yields preliminary directions and the second stage refines them after transforming inputs onto a plane orthogonal to the estimated direction. The method achieves sub-degree angular accuracy, with a mean error around 0.215 degrees on the primary test set and as low as 0.0885 degrees for events with many triggered stations, demonstrating performance comparable to conventional approaches and enabling multimodal analyses across TAIGA detectors. The work highlights the potential of subset-based, two-stage neural pipelines for high-precision EAS direction reconstruction in large detector arrays and sets the stage for integrating data from multiple detector types in TAIGA.

Abstract

The direction of extensive air showers can be used to determine the source of gamma quanta and plays an important role in estimating the energy of the primary particle. The data from an array of non-imaging Cherenkov detector stations HiSCORE in the TAIGA experiment registering the number of photoelectrons and detection time can be used to estimate the shower direction with high accuracy. In this work, we use artificial neural networks trained on Monte Carlo-simulated TAIGA HiSCORE data for gamma quanta to obtain shower direction estimates. The neural networks are multilayer perceptrons with skip connections using partial data from several HiSCORE stations as inputs; composite estimates are derived from multiple individual estimates by the neural networks. We apply a two-stage algorithm in which the direction estimates obtained in the first stage are used to transform the input data and refine the estimates. The mean error of the final estimates is less than 0.25 degrees. The approach will be used for multimodal analysis of the data from several types of detectors used in the TAIGA experiment.

Paper Structure

This paper contains 8 sections, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Two-stage algorithm
  4. Input data
  5. Architecture of the neural networks
  6. Training
  7. Results
  8. Conclusion

Figures (4)

  • Figure 1: The architecute of the first stage neural network.
  • Figure 2: The distribution of the angles between the EAS direction and its estimates by the first stage neural network. (The part where the fractions are indistinguishable from zero is not shown; $\max \omega_1 = 15.135^\circ$, $\max \Omega_1 = 4.714^\circ$.)
  • Figure 3: The distribution of the angles between the EAS direction and its estimates by the second stage neural network. (The part where the fractions are indistinguishable from zero is not shown; $\max \omega_2 = 6.312^\circ$, $\max \Omega_2 = 4.544^\circ$.)
  • Figure 4: The mean estimation errors for events grouped by the number of triggered stations.