Speaker
Description
The identification of $\gamma$-rays from the predominant hadronic-background is a key aspect in their ground-based detection using Imaging Atmospheric Cherenkov Telescopes (IACTs). Current methods rely on Boosted Decision Trees (BDTs) or goodness-of-fit parameters, which are limited in their ability to exploit deep correlations in complex data. Deep learning (DL)-based methods are able to extract such information and offer the potential to improve the telescope and array sensitivities. However, several challenges related to the robustness and applicability of DL-based models remain in the field. A Convolutional and Graph Neural Network (CNN-GNN)-based approach is proposed, which addresses some of these limitations. The model is trained and validated on simulations from the H.E.S.S. experiment, where it demonstrates excellent performance. Preliminary results regarding the model's performance on observational data from the H.E.S.S. experiment are also presented.
