Abstract:Water-Cherenkov-based observatories form the high-duty-cycle and wide field of view backbone for observations of the gamma-ray sky at very high energies. For gamma-ray observations, precise event reconstruction and highly effective background rejection are crucial and have been continuously improving in recent years. In this work, we propose a deep learning application based on graph neural networks (GNNs) for background rejection and energy reconstruction and compare it to state-of-the-art approaches. We find that GNNs outperform hand-designed classification algorithms and observables in background rejection and find an improved energy resolution compared to template-based methods.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: How to improve the performance of gamma - ray observatories based on Water - Cherenkov Detectors (WCDs) in terms of background rejection and energy reconstruction by applying Graph Neural Networks (GNNs). Specifically, the paper focuses on the following aspects: 1. **Background Rejection**: In gamma - ray observations, accurate event reconstruction and efficient background rejection are crucial. Traditional hand - designed classification algorithms and observables have limitations in background rejection. The paper proposes to use GNNs to improve the efficiency of background rejection and compares them with existing methods. 2. **Energy Reconstruction**: Besides background rejection, energy reconstruction is also a key issue in gamma - ray observations. The paper explores the performance of GNNs in energy reconstruction and evaluates their improvement relative to the template method. ### Main Contributions of the Paper - **Innovative Application**: For the first time, graph neural networks are applied to gamma - ray observatories based on water - Cherenkov detectors. - **Performance Improvement**: Research shows that GNNs are significantly superior to existing hand - designed classification algorithms and observables in terms of background rejection, and also have an improvement in energy resolution. - **Data Processing**: Through simulation data sets, the efficiency of GNNs in processing sparse signal patterns is demonstrated, especially in the design of the South - hemisphere Wide - field Gamma - ray Observatory (SWGO). ### Formula Summary The formulas involved in the paper are mainly used to describe the operations of GNNs and performance evaluation metrics: 1. **GNN Operation Formula**: \[ x'_i=\sigma\left(\sum_{j \in N(i)} h_\Theta(x_i \| x_j - x_i)\right) \] where \(x'_i\) is the feature vector of node \(i\) after edge convolution, \(N(i)\) is the set of neighbors of node \(i\), \(h_\Theta\) is the kernel function, and \(\sigma\) is the activation function. 2. **Quality Factor Formula**: \[ Q = \epsilon_\gamma\sqrt{\epsilon_p} \] where \(\epsilon_\gamma\) is the gamma - ray efficiency and \(\epsilon_p\) is the proton contamination rate. 3. **LIC Parameter Formula**: \[ \text{LIC}=\log_{10}\left(\frac{1}{C}\right)=\log_{10}\left(\frac{C_{\text{xPE}40}}{n_{\text{hit}}}\right) \] where \(C_{\text{xPE}40}\) is the maximum charge measured by the PMT at least 40 meters away from the shower core, and \(n_{\text{hit}}\) is the number of water tanks triggered in the event. 4. **PINCness Parameter Formula**: \[ \text{PINCness}=\frac{1}{N}\sum_{i = 0}^{N}\left(\log_{10}(q_i)-\langle\log_{10}(q_i)\rangle\right)^2/\sigma^2 \] where \(q_i\) is the charge of the \(i\)-th triggered PMT, and \(\sigma\) is the uncertainty of the charge. These formulas help readers understand the specific operations and performance evaluation methods of GNNs in processing and analyzing gamma - ray data.

Application of Graph Networks to a wide-field Water-Cherenkov-based Gamma-Ray Observatory

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

Graph Neural Networks for Particle Reconstruction in High Energy Physics detectors

NuGraph2: A Graph Neural Network for Neutrino Physics Event Reconstruction

Graph neural networks at the Large Hadron Collider

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

Comparison of Geometrical Layouts for Next-Generation Large-volume Cherenkov Neutrino Telescopes

Optimizing the Wasserstein GAN for TeV Gamma Ray Detection with VERITAS

Application of Graph Neural Networks in Dark Photon Search with Visible Decays at Future Beam Dump Experiment

Application of Deep Learning Methods for Distinguishing Gamma-Ray Bursts from Fermi/GBM TTE Data

Using Machine Learning to Improve Neutron Identification in Water Cherenkov Detectors

Beyond Grid Data: Exploring Graph Neural Networks for Earth Observation

Analyzing γ rays of the Galactic Center with deep learning

Physics-informed machine learning approaches to reactor antineutrino detection

Deep learning techniques for Imaging Air Cherenkov Telescopes

Survey of Graph Neural Networks and Applications

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

A Graph Neural Network Framework for Grid-Based Simulation

Graph neural networks for reconstruction and classification in KM3NeT

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

Application of Graph Neural Networks and graph descriptors for graph classification