The DPG Spring Meeting in Bonn had to be cancelled! Read more ...
Parts | Days | Selection | Search | Updates | Downloads | Help
T: Fachverband Teilchenphysik
T 5: Machine Learning: QCD and electromagnetic showers
T 5.3: Talk
Monday, March 30, 2020, 17:00–17:15, H-HS III
Generative Models for Fast Shower Simulation — •Sascha Diefenbacher1, Erik Buhmann1, Engin Eren2, Frank Gaede2, and Gregor Kasieczka1 — 1Universität Hamburg, Institute for Experimental Physics — 2Deutsches Elektronen-Synchrotron DESY
In high energy physics, simulations of particle collisions play a vital role in most analysis. A significant portion of the time required for these simulations has to be allocated to modeling how highly energetic particles interact with detectors. These simulation times are bound to increase even further, as increased collider luminosities call for more generated samples and advances in detector technology require these samples to have an increasingly fine resolution. One solution is the use of so called generative machine learning models. These models can learn the properties of a calorimeter shower from a relatively small dataset, and are then able to provide new shower samples orders of magnitude faster than a state of the art, full simulation like Geant4 could. We show results of using the two main generative architectures, Generative Adversarial Networks and Variational AutoEncoders, to generate particle showers in a high granular, 5d calorimeter as proposed by the ILD project.