Distance-Weighted Graph Neural Networks on FPGAs for Real-Time Particle Reconstruction in High Energy Physics

• Main Authors: Iiyama, Yutaro, Cerminara, Gianluca, Gupta, Abhijay, Kieseler, Jan, Loncar, Vladimir, Pierini, Maurizio, Qasim, Shah Rukh, Rieger, Marcel, Summers, Sioni, Van Onsem, Gerrit, Wozniak, Kinga, Ngadiuba, Jennifer, Di Guglielmo, Giuseppe, Duarte, Javier, Harris, Philip, Rankin, Dylan, Jindariani, Sergo, Liu, Mia, Pedro, Kevin, Tran, Nhan, Kreinar, Edward, Wu, Zhenbin
• Format: Journal Article
• Language: English
• Published: 04-02-2021
• Subjects: Computer Science - Learning; Physics - High Energy Physics - Experiment; Physics - Instrumentation and Detectors
• DOI: 10.48550/arxiv.2008.03601

Frontiers in Big Data 3 (2021) 44 Graph neural networks have been shown to achieve excellent performance for
several crucial tasks in particle physics, such as charged particle tracking,
jet tagging, and clustering. An important domain for the application of these
networks is the FGPA-based first layer of real-time data filtering at the CERN
Large Hadron Collider, which has strict latency and resource constraints. We
discuss how to design distance-weighted graph networks that can be executed
with a latency of less than 1$\mu\mathrm{s}$ on an FPGA. To do so, we consider
a representative task associated to particle reconstruction and identification
in a next-generation calorimeter operating at a particle collider. We use a
graph network architecture developed for such purposes, and apply additional
simplifications to match the computing constraints of Level-1 trigger systems,
including weight quantization. Using the $\mathtt{hls4ml}$ library, we convert
the compressed models into firmware to be implemented on an FPGA. Performance
of the synthesized models is presented both in terms of inference accuracy and
resource usage.