Dortmund 2021 – wissenschaftliches Programm

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

T: Fachverband Teilchenphysik

T 65: Experimental methods III

T 65.1: Vortrag

Mittwoch, 17. März 2021, 16:00–16:15, To

Performance of Top-quark Tagging using Unified Flow Object Jets — •Christof Sauer and André Schöning — Physikalisches Institut Universität Heidelberg, Deutschland

Algorithms that identify boosted hadronically decaying objects are used in a variety of physics analyses. The improvements of such taggers have a significant impact on the sensitivity of searches beyond the Standard Model (SM), but also allow for more precise SM measurements. Jets reconstructed from Unified Flow Object (UFO) are going to be the new baseline for large-R(=1.0) jets due to their improved mass and substructure variable resolutions compared to jets built from other objects like, e.g., topological clusters, pFlow or TCCs.
The results shown in this talk are based on the optimization of two different taggers utilizing UFO jets that identify top quarks and reject QCD jets. A simple comparison of various jet substructure variables – in combination with a one-sided mass cut – has been performed to select a list of well discriminating variables for the new UFO jet collection. However, it was shown in earlier analyses that Deep Neural Networks (DNN) for top identification significantly outperform simple cut-based taggers that cut on two to three variables; hence, their usage is the primary consolidated recommendation for the upcoming release.
Finally, a comparison of both taggers (cut- and DNN-based) using UFOs as well as their performance improvement with respect to the previous baseline, i.e. LCTopo, will be presented along with an outlook regarding upcoming tagger concepts involving UFO jets.