Aachen 2019 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
T: Fachverband Teilchenphysik
T 4: Deep Learning I
T 4.8: Vortrag
Montag, 25. März 2019, 17:45–18:00, H06
Precise simulation of electromagnetic calorimeter showers using a Wasserstein Generative Adversarial Network — Martin Erdmann1, Jonas Glombitza1, and •Thorben Quast1,2 — 1Physikalisches Institut 3A, RWTH Aachen — 2EP-LCD, CERN
The increased instantaneous luminosity at the High Luminosity LHC will raise the computing requirements for event reconstruction and analysis for current LHC-based experiments, hence limiting the available resources for the simulation of particles traversing matter. Developments on the performance of state-of-the-art simulation frameworks such as Geant4 are proceeding but are unlikely to fully compensate for this trend. Generative adversarial neural networks (GANs) have been shown to provide promising fast simulation models. Wasserstein GANs (WGANs) are a variant of this method. They employ a more robust metric for the adversarial training of the generator network. In this talk, we show our adaptation of the WGAN concept for the generation of electromagnetic showers inside a realistic setup of a multi-layer sampling calorimeter. In addition, conditioning on the energy of the incident particle and on its impact position is integrated through two auxiliary regression networks. Overall, the quality of these fast shower simulations with the WGAN reaches the level of showers generated with the GEANT4 program in most aspects. At the same time, the computational speed-up compared to traditional sequential simulations amounts to several orders of magnitudes.