Bonn 2020 – wissenschaftliches Programm

Die DPG-Frühjahrstagung in Bonn musste abgesagt werden! Lesen Sie mehr ...

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

T: Fachverband Teilchenphysik

T 5: Machine Learning: QCD and electromagnetic showers

T 5.5: Vortrag

Montag, 30. März 2020, 17:30–17:45, H-HS III

Studies on using Generative Adversarial Networks to simulate parton showers — Johannes Erdmann, •Alexander Froch, and Olaf Nackenhorst — TU Dortmund, Lehrstuhl für Experimentelle Physik IV

Monte Carlo (MC) simulations are one of the basic instruments in data analysis of high energy physics experiments. The three main parts that need to be simulated are the hard scattering process, the parton shower + hadronisation and the detector simulation. Although MC simulations bring great benefits for the data analysis of high energy physics experiments, the costs and time needed to produce them are significantly high. Generative Adversarial Networks (GANs) can be trained with samples from MC simulations to be used for fast MC simulations due to their characteristic as a much less computing-intensive model. It has been shown that GANs are capable to simulate the hard scattering process or imitate even the whole MC simulation process. They were also used as a fast detector simulation trained on samples generated with the GEANT4 detector simulation. Significantly reduced computing times for the event generation were accomplished in comparison to the GEANT4 detector simulation. Motivated by these results we examine the feasibility of generating key features of the parton shower with GANs. In this presentation the latest status of our studies is shown.