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

TL;DR

This paper presents the design and implementation of distance-weighted graph neural networks optimized for FPGA hardware, achieving real-time particle reconstruction with less than 1 microsecond latency for high energy physics applications.

Contribution

It introduces a novel FPGA-compatible graph neural network architecture with quantization and simplifications tailored for real-time particle physics data filtering.

Findings

Achieves sub-microsecond inference latency on FPGA

Maintains high accuracy in particle reconstruction tasks

Optimizes resource usage for FPGA deployment

Abstract

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.