Speaker
Description
Accurate reconstruction of Ultra-High-Energy Cosmic Ray (UHECR) parameters is crucial for understanding their origins and composition. We present a newly developed Deep Neural Network (DNN) approach based on the AixNet architecture for reconstructing UHECR parameters from Telescope Array surface detector (SD) data. This model reconstructs key parameters, including energy, arrival direction, core position, $X_{\text{max}}$, and primary mass, by analyzing time traces and spatial correlations. Monte Carlo simulations for four mass groups (proton, helium, CNO, and iron) demonstrate that the DNN improves the resolution of energy, direction, and core position compared to standard reconstruction methods. We expect that the DNN will achieve these improvements with looser data quality requirements, potentially increasing the available event statistics. We provide expected resolution figures and systematic studies from simulations and validate the DNN’s performance using hybrid data.
