Dortmund 2021 – wissenschaftliches Programm

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

T: Fachverband Teilchenphysik

T 21: Data analysis, Information technology I

T 21.3: Vortrag

Montag, 15. März 2021, 16:30–16:45, Tu

Utilization of GPUs in the training of neural networks — Günter Quast, Roger Wolf, Janek Bechtel, Sebastian Brommer, Rene Caspart, Ralf Schmieder, Felix Heyen, Gessi Risto, Andrew Issac, and •Tim Voigtländer — Karlsruher Institut für Technologie, Karlsruhe, Deutschland

Machine learning has become a commonly used technique in recent particle physics analyses. As machine learning algorithms become more refined and the neural networks used in recent analyses have become more complex, the usage of specialized computing resources needs to be explored, in order to ensure a reasonable turnaround cycle, especially for the training of the networks. In this talk the utilization of a cluster of GPUs used to accelerate the training process is shown by the example of the multilayer neural networks used in the CMS Higgs → ττ analysis.