Three Dimensional Energy Parametrized Generative Adversarial Networks for Electromagnetic Shower Simulation

High Energy Physics (HEP) simulations are traditionally based on the Monte Carlo approach and generally rely on time consuming calculations. The present work investigates the use of Generative Adversarial Networks (GANs) as a fast alternative. Our approach treats the energy deposited by a particle i...

Full description

Saved in:
Bibliographic Details
Published in2018 25th IEEE International Conference on Image Processing (ICIP) pp. 3913 - 3917
Main Authors Khattak, Gul rukh, Vallecorsa, Sofia, Carminati, Federico
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.10.2018
Subjects
Deep Neural Networks
Detectors
Gallium nitride
GAN
Generative adversarial networks
Generators
HEP
Monte Carlo methods
Physics
Simulation
Training
Online AccessGet full text

Cover

More Information
Summary:High Energy Physics (HEP) simulations are traditionally based on the Monte Carlo approach and generally rely on time consuming calculations. The present work investigates the use of Generative Adversarial Networks (GANs) as a fast alternative. Our approach treats the energy deposited by a particle inside a calorimeter detector as a three-dimensional image. True three-dimensional convolutions can be employed to capture the spatio-temporal correlation of shower energy depositions. Three-dimensional images are generated, conditioned on the energy of the incoming particle and validated against Monte Carlo simulation. The results show an agreement to full Mote Carlo simulations well within 10% thus proving that GAN can be used as a fast alternative for simulation of HEP detector response.
ISSN:2381-8549
DOI:10.1109/ICIP.2018.8451587