Graph Neural Network-Based Track Finding in the LHCb Vertex Detector

TL;DR

This paper introduces a GNN-based track-finding pipeline for the LHCb vertex detector, demonstrating its efficiency and performance improvements over classical methods on GPU architectures, with a focus on batched execution and neural network quantization.

Contribution

It presents a novel batched GNN tracking pipeline for high-energy physics, optimized for GPU, and evaluates its performance and physics accuracy against traditional algorithms.

Findings

GNN pipeline achieves comparable physics performance to classical algorithms.

Batched execution significantly improves GPU computational efficiency.

Neural network quantization maintains physics accuracy while reducing computational cost.

Abstract

The next decade will see an order of magnitude increase in data collected by high-energy physics experiments, driven by the High-Luminosity LHC (HL-LHC). The reconstruction of charged particle trajectories (tracks) has always been a critical part of offline data processing pipelines. The complexity of HL-LHC data will however increasingly mandate track finding in all stages of an experiment's real-time processing. This paper presents a GNN-based track-finding pipeline tailored for the Run 3 LHCb experiment's vertex detector and benchmarks its physics performance and computational cost against existing classical algorithms on GPU architectures. A novelty of our work compared to existing GNN tracking pipelines is batched execution, in which the GPU evaluates the pipeline on hundreds of events in parallel. We evaluate the impact of neural-network quantisation on physics and computational performance, and comment on the outlook for GNN tracking algorithms for other parts of the LHCb track-finding pipeline.