Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

HK: Fachverband Physik der Hadronen und Kerne

HK 27: Poster Session

HK 27.17: Poster

Mittwoch, 18. März 2026, 16:15–18:30, Redoutensaal

Particle Track classification using Object Condensation at MAGIX — •Nils Hesse for the MAGIX collaboration — JGU Mainz

Machine Learning has been a part of modern physics for many years now, aiding with denoising, particle identification and even being used for generating particle events. However, a general limitation lies in the fact that neural networks have a fixed number of inputs and outputs, making classification of an unknown number of particle tracks not just more complex, but also inefficient.

Object Condensation is a loss function developed by Jan Kieseler in 2020, it solves this problem by transforming the input data into a learned condensation space and clustering in said space. This work explores the application of Object Condensation to the Prototype TPC of the MAinz Gas Injection target eXperiment (MAGIX) at the Mainz Energy-Recovering Superconducting Accelerator (MESA). Preliminary results are presented alongside an overview of the underlying principles of the Object Condensation approach.

Keywords: MAGIX; Machine Learning; Object Condensation; TPC; MESA