Abstract:For the past year, the <a class="link-external link-http" href="http://HEP.TrkX" rel="external noopener nofollow">this http URL</a> project has been investigating machine learning solutions to LHC particle track reconstruction problems. A variety of models were studied that drew inspiration from computer vision applications and operated on an image-like representation of tracking detector data. While these approaches have shown some promise, image-based methods face challenges in scaling up to realistic HL-LHC data due to high dimensionality and sparsity. In contrast, models that can operate on the spacepoint representation of track measurements ("hits") can exploit the structure of the data to solve tasks efficiently. In this paper we will show two sets of new deep learning models for reconstructing tracks using space-point data arranged as sequences or connected graphs. In the first set of models, Recurrent Neural Networks (RNNs) are used to extrapolate, build, and evaluate track candidates akin to Kalman Filter algorithms. Such models can express their own uncertainty when trained with an appropriate likelihood loss function. The second set of models use Graph Neural Networks (GNNs) for the tasks of hit classification and segment classification. These models read a graph of connected hits and compute features on the nodes and edges. They adaptively learn which hit connections are important and which are spurious. The models are scaleable with simple architecture and relatively few parameters. Results for all models will be presented on ACTS generic detector simulated data.

Deep learning for track recognition in pixel and strip-based particle detectors

Novel deep learning methods for track reconstruction

Particle Identification in Particle Tracking Velocimetry Using Two-Stage Neural Networks

Track finding with deep neural networks

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

Particle Track Reconstruction using Geometric Deep Learning

Charged Particle Tracking via Edge-Classifying Interaction Networks

End-to-End Multi-Track Reconstruction using Graph Neural Networks at Belle II

Graph Neural Networks for Particle Reconstruction in High Energy Physics detectors

EggNet: An Evolving Graph-based Graph Attention Network for Particle Track Reconstruction

An Object Condensation Pipeline for Charged Particle Tracking at the High Luminosity LHC

Track Seed Classification with Deep Neural Networks

Performance of a Geometric Deep Learning Pipeline for HL-LHC Particle Tracking

Machine Learning for Track Finding at PANDA

Deep Learning-Based Spatiotemporal Multi-Event Reconstruction for Delay Line Detectors

Real-Time 3D Tracking of Multi-Particle in the Wide-Field Illumination Based on Deep Learning

A Deep Learning Approach to Multi-Track Location and Orientation in Gaseous Drift Chambers

Track Reconstruction using Geometric Deep Learning in the Straw Tube Tracker (STT) at the PANDA Experiment

HyperTrack: Neural Combinatorics for High Energy Physics

Graph Neural Networks for Charged Particle Tracking on FPGAs

NuGraph2: A Graph Neural Network for Neutrino Physics Event Reconstruction