Fast and Precise Learned Charged-Particle Trajectory Regression at the Large Hadron Collider
Abstract
We propose a training recipe that treats charged-particle trajectory parameter regression on high-energy physics detector data as a sequence-modeling task. Kalman filters and linearized least-squares fits have been the classical standard approach for this task: they are optimal estimators for sparsely sampled linear-Gaussian data and are commonly used for trajectory parameter regression (fitting). The classical fitting techniques implemented for this domain reach a final precision of one part in 105 through detailed modeling of detector geometry, material, detection effects and precise numerical integration of the equations of motion through the detector's inhomogeneous magnetic field. With this study, we demonstrate that using a bidirectional gated linear recurrent encoder, one is able to reproduce the full precision of classical track fitting techniques. Using a custom kernel, we also achieve significantly higher throughput during GPU inference, compared to classical fitting software running on similarly priced multi-core CPU servers representing typically employed hardware. Such a speedup would lead to considerable cost savings for the pattern recognition at the Large Hadron Collider. To our knowledge, this is the first end-to-end learned track fit to reach the full precision and, at the same time, offer the opportunity to reduce the computing costs.
Related
- Machine-learned particle flow as a foundation model for collider physics
- Smite: A quasiclassical trajectory (QCT) program for bimolecular collisions and unimolecular dynamics on ab-initio, and machine-learned potential energy surfaces
- Particle track reconstruction in high-density collider environments with an Evolving-Geometry Transformer
- A neural delta-f particle-in-cell method: evolving the background with learned backward flows
- Direct Trajectory Reconstruction for Fast 3D X-ray Particle Tracking Velocimetry in Porous Media
BibTeX
@misc{renusch2026fast,
title = {{Fast and Precise Learned Charged-Particle Trajectory Regression at the Large Hadron Collider}},
author = {Jonathan Renusch and Benjamin Huth and Daniel Murnane and Eleni Xochelli and Doğa Elitez and Paul Gessinger-Befurt and Andreas Stefl and Jeremy Couthures and Andreas Salzburger and Lukas Heinrich and Michael Kagan and Markus Elsing},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.32479},
eprint = {2609.32479},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.32479},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Fast and Precise Learned Charged-Particle Trajectory Regression at the Large Hadron Collider
AU - Jonathan Renusch
AU - Benjamin Huth
AU - Daniel Murnane
AU - Eleni Xochelli
AU - Doğa Elitez
AU - Paul Gessinger-Befurt
AU - Andreas Stefl
AU - Jeremy Couthures
AU - Andreas Salzburger
AU - Lukas Heinrich
AU - Michael Kagan
AU - Markus Elsing
PY - 2026
DA - 2026/09/26
PB - arXiv
DO - 10.48550/arxiv.2609.32479
UR - https://arxiv.org/abs/2609.32479
AB - We propose a training recipe that treats charged-particle trajectory parameter regression on high-energy physics detector data as a sequence-modeling task. Kalman filters and linearized least-squares fits have been the classical standard approach for this task: they are optimal estimators for sparsely sampled linear-Gaussian data and are commonly used for trajectory parameter regression (fitting). The classical fitting techniques implemented for this domain reach a final precision of one part in 10^5 through detailed modeling of detector geometry, material, detection effects and precise numerical integration of the equations of motion through the detector's inhomogeneous magnetic field. With this study, we demonstrate that using a bidirectional gated linear recurrent encoder, one is able to reproduce the full precision of classical track fitting techniques. Using a custom kernel, we also achieve significantly higher throughput during GPU inference, compared to classical fitting software running on similarly priced multi-core CPU servers representing typically employed hardware. Such a speedup would lead to considerable cost savings for the pattern recognition at the Large Hadron Collider. To our knowledge, this is the first end-to-end learned track fit to reach the full precision and, at the same time, offer the opportunity to reduce the computing costs.
ER -