Analysis of the TAIGA-HiSCORE Data Using the Latent Space of Autoencoders

TL;DR

This work investigates reconstructing the energy of extensive air showers using the latent space of a convolutional autoencoder trained on TAIGA-HiSCORE quasi-images. By replacing hand-crafted auxiliary parameters with latent features, the authors create a data-driven, potentially multimodal pipeline in which an energy-predicting neural network operates on the encoded representation. They demonstrate that a latent space of $N=12$ achieves a favorable trade-off between reconstruction quality and model complexity, attaining energy-reconstruction errors on the order of 13–24% across TeV–PeV energies, comparable to conventional methods. The approach holds promise for seamless integration of heterogeneous TAIGA data streams and paves the way for joint analyses with TAIGA-IACTs, while enabling future expansion to additional EAS parameters and primary particle types.

Abstract

The aim of extensive air shower (EAS) analysis is to reconstruct the physical parameters of the primary particle that initiated the shower. The TAIGA experiment is a hybrid detector system that combines several imaging atmospheric Cherenkov telescopes (IACTs) and an array of non-imaging Cherenkov detectors (TAIGA-HiSCORE) for EAS detection. Because the signals recorded by different detector types differ in physical nature, the direct merging of data is unfeasible, which complicates multimodal analysis. Currently, to analyze data from the IACTs and TAIGA-HiSCORE, a set of auxiliary parameters specific to each detector type is calculated from the recorded signals. These parameters are chosen empirically, so there is no certainty that they retain all important information and are the best suited for the respective problems. We propose to use autoencoders (AE) for the analysis of TAIGA experimental data and replace the conventionally used auxiliary parameters with the parameters of the AE latent space. The advantage of the AE latent space parameters is that they preserve essential physics from experimental data without prior assumptions. This approach also holds potential for enabling seamless integration of heterogeneous IACT and HiSCORE data through a joint latent space. To reconstruct the parameters of the primary particle of the EAS from the latent space of the AE, a separate artificial neural network is used. In this paper, the proposed approach is used to reconstruct the energy of the EAS primary particles based on Monte Carlo simulation data for TAIGA-HiSCORE. The dependence of the energy determination accuracy on the dimensionality of the latent space is analyzed, and these results are also compared with the results obtained by the conventional technique. It is shown that when using the AE latent space, the energy of the primary particle is reconstructed with satisfactory accuracy.

Figures (7)

• Figure 1: A visualization of a TAIGA-HiSCORE event. Each point corresponds to a HiSCORE station, the color from purple to yellow indicates the time of signal recording, the size indicates the signal amplitude, and small gray dots indicate stations that were not triggered.
• Figure 2: The general architecture of the proposed model. First, the AE is trained to reconstruct the registered TAIGA-HiSCORE data, then the energy reconstruction network is trained to reconstruct the energy from the AE latent space.
• Figure 3: The architecture of the encoder.
• Figure 4: Image reconstruction results for three events with different numbers of triggered stations. The original quasi-image is shown on the left, and its corresponding reconstructed image is on the right. Each image pair corresponds to a single event. The number of triggered stations is 20 for the first event, 40 for the second, and 80 for the third.
• Figure 5: The architecture of the network for energy reconstruction.