EVENTI
Lunedì 1 dicembre alle 14:30, in Aula seminari, la dott.ssa Federica Cuna (INFN Bari) terrà un seminario dal titolo:
Advanced Tracking Analysis in Space Experiments with Graph Neural Networks
Abstract:
The integration of advanced artificial intelligence (AI) techniques into astroparticle physics represents a transformative shift in how data analysis and experimental design are approached. As space-based missions become increasingly complex, leveraging AI has become essential for optimizing performance and achieving robust scientific results.
We present the development of innovative AI-driven algorithms for particle tracking, focusing on the use of Graph Neural Networks (GNNs)—a class of geometric deep learning models particularly well-suited to the graph-like structure of tracking systems.
One of the major challenges in space-based tracking systems is the high-noise environment, especially due to backscattered tracks from calorimeters, which obscure the identification of the primary particle trajectory. To address this, we propose a GNN-based node classification algorithm capable of distinguishing signal hits from backscattering hits, thereby enabling the accurate reconstruction of particle tracks.
Our approach has been applied on simulated data from a real astroparticle experiment, ensuring that the algorithm is tailored to realistic detector conditions and configurations. The model effectively identifies the hits associated with the primary particle and enables the reconstruction of key track parameters with high fidelity.
By tackling the intrinsic complexity of the space detector environment using cutting-edge AI methodologies, this work contributes to improving the precision and accuracy of data analysis in astroparticle physics, ultimately supporting more insightful scientific discoveries.
Ref. Prof. F. de Palma