Description
The integration of advanced artificial intelligence (AI) techniques in astroparticle experiments marks a significant breakthrough in data analysis and experimental methodology. As space missions grow in complexity, AI-driven approaches are crucial for optimizing performance and maximizing scientific output. In this study, we present a fully custom-designed Transformer-based model specifically developed for calorimeters in space-based experiments. A key objective of these experiments is the precise classification of particle types, such as electrons and protons. By leveraging Transformers’ ability to capture complex dependencies across thousands of channels and high-dimensional feature spaces, our approach enhances classification accuracy and robustness. Additionally, we implement GPU-based parallelization techniques to efficiently handle large-scale data processing, significantly accelerating model training and inference. This methodology not only improves data interpretation in astroparticle physics but also extends classification capabilities across an extensive energy range, from 1 GeV to 100 TeV.
