SMuK 2023 – wissenschaftliches Programm

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

T: Fachverband Teilchenphysik

T 22: Calorimeter / Detector Systems I

T 22.1: Vortrag

Montag, 20. März 2023, 16:30–16:45, WIL/C133

Bitwise Optimization of Artificial Neural Networks for the Energy Reconstruction of ATLAS Liquid-Argon Calorimeter Signals — •Alexander Lettau, Anne-Sophie Berthold, Nick Fritzsche, Christian Gutsche, Arno Straessner, and Johann Christoph Voigt — Institut für Kern- und Teilchenphysik, Dresden, Deutschland

The LHC will be upgraded to become the High-Luminosity-LHC, with significantly increased numbers of simultaneous particle collisions. With this upgrade, up to 200 pile-up events are expected within one bunch crossing. To cope with that, processing of the signals of the Liquid-Argon Calorimeter will need to be improved, because conventional algorithms are expected to lose performance. Artifical neural networks provide one way to deal with this. It has been shown, that convolutional neural networks are able to detect signals and reconstruct their energy with good performance. These networks are planned to be executed on Field Programmable Gate Arrays (FPGA) which have limited resources in signal, processing units, logic and memory. This talk will deal with the quantization of neural networks, a technique to reduce the resources needed for neural networks, by reducing the precision of the weights, biases and activations, while keeping the performance.