One of the major challenges in the Large Hadron Collider forward (LHCf) experiment is the accurate reconstruction of calorimetric clusters when multiple particles hit the same detector tower simultaneously. Traditional reconstruction methods struggle with overlapping signals, especially in events involving more than two particles or a combination of photons and neutrons. This paper presents the development of machine learning (ML) techniques to improve the reconstruction efficiency of such complex events. We discuss the motivations for integrating ML into the LHCf reconstruction pipeline, outline the ML approach and dataset preparation, and compare the performance of ML models with standard methods. The results demonstrate a significant improvement in reconstructing multi-hit events, which is essential for analyses involving π0, η, Ks0 mesons, and Λ0 baryon. Finally, we explore future prospects for ML applications in the LHCf experiment.
Reconstruction of multiple calorimetric clusters in the LHCf experiment with machine learning techniques / Piparo G.; Adriani O.; Berti E.; Betti P.; Bonechi L.; Bongi M.; D'Alessandro R.; Detti S.; Gensini E.; Haguenauer M.; Isseverc C.; Itow Y.; Kasahara K.; Kinoshita K.; Kobayashi H.; Leitgeb C.; Matsubara Y.; Menjo H.; Muraki Y.; Papini P.; Piparo G.; Ricciarini S.; Sako T.; Sakuma M.; Sakurai N.; Scaringella M.; Shimizu Y.; Tamura T.; Tiberio A.; Torii S.; Tricomi A.; Turner W.C.; Yoshida K.. - In: POS PROCEEDINGS OF SCIENCE. - ISSN 1824-8039. - ELETTRONICO. - 476:(2025), pp. 1016.0-1016.0.
Reconstruction of multiple calorimetric clusters in the LHCf experiment with machine learning techniques
Adriani O.;Betti P.;Bongi M.;D'Alessandro R.;Gensini E.;Tiberio A.;
2025
Abstract
One of the major challenges in the Large Hadron Collider forward (LHCf) experiment is the accurate reconstruction of calorimetric clusters when multiple particles hit the same detector tower simultaneously. Traditional reconstruction methods struggle with overlapping signals, especially in events involving more than two particles or a combination of photons and neutrons. This paper presents the development of machine learning (ML) techniques to improve the reconstruction efficiency of such complex events. We discuss the motivations for integrating ML into the LHCf reconstruction pipeline, outline the ML approach and dataset preparation, and compare the performance of ML models with standard methods. The results demonstrate a significant improvement in reconstructing multi-hit events, which is essential for analyses involving π0, η, Ks0 mesons, and Λ0 baryon. Finally, we explore future prospects for ML applications in the LHCf experiment.
I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
