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

T: Fachverband Teilchenphysik

T 30: Top Physics II (Properties)

T 30.2: Vortrag

Dienstag, 1. April 2025, 16:30–16:45, VG 1.103

Optimizing Jet-Parton Assignments in Fully Hadronic Top-Quark Decays: A Comparison of SPANet and Traditional Methods — •Nico Rehberg, Johannes Lange, Hartmut Stadie, and Peter Schleper — Institute of Experimental Physics, Hamburg University, Germany

Accurate jet-parton assignments in fully hadronic top-quark decays are crucial for the precise reconstruction of the top-quark mass. Traditional approaches, such as applying a kinematic fit, provide reliable results but are limited by the rapid increase in possible permutations as the number of jets grows. These methods become less efficient due to combinatorics in case of high jet multiplicities, and they do not make use of dynamical properties of tt processes. The Symmetry Preserving Attention Network (SPANet), a machine learning-based approach, addresses these challenges by exploiting the inherent symmetries of the assignment problem, resulting in improved scaling during inference. This talk provides a brief overview of the network’s structure and presents a comparison of assignment results between SPANet and traditional approaches, including the χ²-method and kinematic fit.