Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
T: Fachverband Teilchenphysik
T 21: Experimentelle Methoden 1 (Computing, Machine Learning, Statistik)
T 21.3: Vortrag
Montag, 27. März 2017, 17:15–17:30, JUR 253
First steps towards an improved tuning method for Monte Carlo generators — •Fabian Klimpel1,2, Stefan Kluth1, and Andrea Knue1 — 1Max Planck Institut fuer Physik, Munich — 2Technical University Munich
In high energy physics, Monte Carlo (MC) generators are used for the simulation of physics processes. In the simulation, parameters in the hard interaction and in the parton shower can be varied in a well defined range to achieve a better description of the data distributions (MC tuning). To do a full tuning, several parameters are varied and each parameter set leads to a simulated sample which is compared to the data. A binwise parametrization of the parameter variations is performed using the "Professor 2.4" framework. These functions are then optimized with respect to the measured data which are provided in the "Rivet" framework. This optimization should deliver the requested parameter values. In this talk an investigation of the stability of the fixed order polynomial interpolation performed by "Professor 2.4" is presented. This will be shown in comparison to a binwise adaptive fitting method. The optimization performed by "Professor 2.4" will be compared to a tuning performed using the Bayesian Analysis Toolkit (BAT).