About
HEP advanced tracking algorithms at the exascale (Project Exa.TrkX)
Summary
The proposed work is focused on developing algorithms targeted towards scaling the training and inference of Graph Neural Networks (GNNs) for edge classification in the context of particle track finding for High-Energy Physics (HEP). Track reconstruction (tracking) connects the measured 3D points (hits) to reconstruct charged particle trajectories. Accurately identifying particle tracks from high-density particle hits datasets is essential for the next generation of HEP experiments identified as priorities by the P5 panel. However, the computational complexity of traditional tracking algorithms grows faster than linear with the number of tracking particles, making particle tracking a limiting factor, for example, for HL-LHC (High-Luminosity Large Hadron Collider) physics performance.
Contributors
Collaborators
Exa.TrkX is a follow-up to the the HEP.TrkX pilot project. It relies on the aCTS toolkit to simulate a generic HL-LHC detector, and more recently to benchmark the performance of its models. Exa.TrkX is collaborating with the FastML Lab to deploy GNN models on FPGA systems. Exa.TrkX is also collaborating with the NERSC Big Data center, and the Exalearn co-design center to demonstrate distributed training and model hyperparameter optimization at scale on HPC systems.
Publications
- Performance of a Geometric Deep Learning Pipeline for HL-LHC Particle Tracking (Associated Code ) Eur. Phys. J. C 81, 876 (2021)
- CTD 2023
- ACAT 2024
Conference Contributions
- Hierarchical Graph Neural Networks for Particle Track Reconstruction Presented at ACAT 2022 (Associated Code ).
- ATLAS ITk Track Reconstruction with a GNN-based pipeline Presented at Connecting the Dots 2022 (Associated Code).
- Accelerating the Inference Time of Machine Learning-based Track Finding Pipeline Presented at ACAT 2021 ( Associated Code ).
- Graph Neural Network for Large Radius Tracking Presented at ACAT 2021
- Graph Neural Network for Object Reconstruction in Liquid Argon Time Projection Chambers Presented at CHEP 2021
- Distributed Training of GNNs on HPCs Presented at the 4th Inter-experiment Machine Learning Workshop ( Associated Code ).
- "Track Seeding and Labelling with Embedded-space Graph Neural Networks". Presented at Connecting the Dots 2020 - ( Associated Code ).
- "Graph Neural Networks for Particle Reconstruction in High Energy Physics Detectors". Presented at NeurIPS 2019 Workshop "Machine Learning and the Physical Sciences" - (NeurIPS Poster) ( Associated Code ).
Presentations
- Full-length tutorial on Tracking with Graph Neural Networks (Sep 2023, Heidelberg) Part 1 - Part 2
- Graph Neural Networks for High Luminosity Track Reconstruction (CERN EP-IT Data science seminar).
- Graph Neural Networks for Reconstruction in DUNE (presented at the Dec 4th CLARIPHY topical meeting).
- Tracking with GNNs (in-depth code walk-through at the 4th Inter-experiment Machine Learning Workshop) (colab notebook) ( Associated Code ).
- Graph Neural Networks for Particle Tracking (A non-specialist introduction to Exa.TrkX tracking models).