Abstract:Imaging Atmospheric Cherenkov Telescopes (IACTs) of gamma ray observatory TAIGA detect the Extesnive Air Showers (EASs) originating from the cosmic or gamma rays interactions with the atmosphere. Thereby, telescopes obtain images of the EASs. The ability to segregate gamma rays images from the hadronic cosmic ray background is one of the main features of this type of detectors. However, in actual IACT observations simultaneous observation of the background and the source of gamma ray is needed. This observation mode (called wobbling) modifies images of events, which affects the quality of selection by neural networks.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: **How to effectively distinguish gamma - ray events from hadronic cosmic - ray background events in the images obtained from Imaging Atmospheric Cherenkov Telescopes (IACTs) by using deep - learning methods, especially the impact on image recognition quality when using the "wobbling pointing mode" for observation**. ### Specific problem description: 1. **Distinction between gamma - ray and hadronic background**: - Imaging Atmospheric Cherenkov Telescopes (IACTs) are used to detect Extended Air Showers (EAS) produced by the interaction of high - energy gamma - rays or cosmic - rays with the atmosphere. These showers form images on the detector, and distinguishing gamma - ray events from hadronic cosmic - ray background events is a key task of IACTs. 2. **Impact of the wobbling mode**: - In actual observations, IACTs usually adopt the "wobbling pointing mode", that is, the gamma - ray source is not at the center of the camera but at a certain distance away. This mode will cause the image to be displaced, thus affecting the classification effect of neural networks on events. 3. **Data imbalance problem**: - In the experiment, the ratio of gamma - ray events to hadronic cosmic - ray events is very unbalanced (for example, 1:10000), which poses higher requirements for the classification model. ### Research objectives: - **Develop deep - learning methods** to improve the accuracy of TAIGA - IACT data processing and analysis and partially realize automation. - **Evaluate the performance of different neural network architectures** (such as Convolutional Neural Network (CNN), fully - connected neural network and hybrid networks combining the two) in distinguishing gamma - rays and hadronic background. - **Consider the impact of the wobbling mode on the image** and propose corresponding image enhancement methods to ensure the unity between simulated data and experimental data. Through the above research, the author hopes to improve the existing methods so that deep learning can more effectively identify gamma - ray signals in actual astronomical observations.

Selection of gamma events from IACT images with deep learning methods

Using deep learning methods for IACT data analysis in gamma-ray astronomy: A review

Exploring deep learning as an event classification method for the Cherenkov Telescope Array

Deep learning techniques for Imaging Air Cherenkov Telescopes

Extracting gamma-ray information from images with convolutional neural network methods on simulated Cherenkov Telescope Array data

Signal-background separation and energy reconstruction of gamma rays using pattern spectra and convolutional neural networks for the Small-Sized Telescopes of the Cherenkov Telescope Array

Dimensionality reduction and sensitivity improvement for TACTIC Cherenkov data using t-SNE machine learning algorithm

A new method of reconstructing images of gamma-ray telescopes applied to the LST-1 of CTAO

A background-estimation technique for the detection of extended gamma-ray structures with IACTs

Primary gamma ray selection in a hybrid timing/imaging Cherenkov array

Event Selection and Background Rejection in Time Projection Chambers Using Convolutional Neural Networks and a Specific Application to the AdEPT Gamma-ray Polarimeter Mission

Application of Graph Networks to background rejection in Imaging Air Cherenkov Telescopes

A Hybrid Approach to Event Reconstruction for Atmospheric Cherenkov Telescopes Combining Machine Learning and Likelihood Fitting

Application of Deep Learning Methods Combined with Physical Background in Wide Field of View Imaging Atmospheric Cherenkov Telescopes

Deep Learning and IACT: Bridging the gap between Monte-Carlo simulations and LST-1 data using domain adaptation

A Systematic Assessment of Data Volume Reduction for IACTs

Ultra-fast generation of air shower images for Imaging Air Cherenkov Telescopes using Generative Adversarial Networks

Using a Conditional Generative Adversarial Network to Control the Statistical Characteristics of Generated Images for IACT Data Analysis

An Approach to Gamma-Hadron Separation via Measurements of the Middle-UV Fraction of EAS emission by an Imaging Atmospheric Cherenkov Telescope with UV-Sensitive SiPM detectors