Abstract

The Tunka Advanced Instrument for gamma-ray and cosmic ray Astrophysics (TAIGA) is a hybrid observatory for the detection of extensive air showers (EAS), produced by high-energy gamma rays and cosmic rays. The complex consists of such facilities as TAIGA-IACT, TAIGA-HiSCORE, and a variety of others. The goal of the study is to introduce a deep learning-based technique for EAS axis reconstruction. A convolutional neural network (CNN) model is proposed, while HiSCORE events, consisting of time-amplitude data, are treated as images by the model. Reasoning behind the CNN model and model efficacy will be discussed, along with [preliminary] results for EAS axis direction determination. This article will show that the accuracy of the model reaches 1\({}^{\circ}\)–2\({}^{\circ}\) for the zenith and azimuthal angles, however, the accuracy of the model does not reach the accuracy of conventional methods.

中文翻译：

---

卷积神经网络在TAIGA-HiSCORE实验中数据分析的应用

摘要

Tunka 先进伽马射线和宇宙射线天体物理学仪器 (TAIGA) 是一个混合观测站，用于探测由高能伽马射线和宇宙射线产生的大范围空气簇射 (EAS)。该综合体由 TAIGA-IACT、TAIGA-HiSCORE 等设施组成。该研究的目标是引入一种基于深度学习的 EAS 轴重建技术。提出了卷积神经网络（CNN）模型，而由时间幅度数据组成的 HiSCORE 事件被该模型视为图像。将讨论 CNN 模型背后的推理和模型功效，以及 EAS 轴方向确定的[初步]结果。本文将展示模型对于天顶角和方位角的精度达到 1 \({}^{\circ}\) –2 \({}^{\circ}\) ，但是模型的精度达不到常规方法的准确度。