Ponente
Descripción
The Cherenkov Telescope Array Observatory (CTAO) will be the next generation of very high energy (VHE) gamma-ray observatories, using Imaging Atmospheric Cherenkov Telescopes (IACTs) such as the LST-1 (Large-Sized-Telescope-1). In this contribution, we present a successful application of deep learning techniques for event reconstruction in real LST-1 data, using the CTLearn framework. The model is based on convolutional neural networks (CNNs) applied to calibrated images, allowing the precise estimation of physical event parameters such as energy, direction and gammaness (a score used in the selection of gamma-ray events over background).
The results show a significant improvement in the differential sensitivity of LST-1 in the low energy range up to about 500 GeV, compared to classical methods based on Random Forests. This improvement is particularly reflected in better angular resolution, energy resolution, and improved the gammaness estimation, which directly contributes to the increase in sensitivity. The work demonstrates the potential of deep learning as a robust and effective tool for the analysis of real data in IACT telescopes and reinforces its applicability in operational scientific environments.
