Speaker
Description
LHAASO-AI explores deep-learning-based approaches for event reconstruction and classification in LHAASO. We have developed models for four major tasks: gamma/hadron discrimination, arrival-direction reconstruction, shower-core reconstruction, and energy reconstruction. The models are trained and systematically evaluated using large-scale Monte Carlo simulations. Validation with experimental data has been established for gamma/hadron discrimination and directional reconstruction, while corresponding studies for shower-core and energy reconstruction are still in progress. For KM2A, the inference procedure has been optimized for computational efficiency, and an AI-based reconstruction pipeline has been implemented on the IHEP computing cluster. These studies demonstrate the potential of AI-based methods to improve the reconstruction performance and analysis capability of LHAASO.