skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Novel deep learning methods for track reconstruction

Abstract

For the past year, the HEP.TrkX 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 whichmore » 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.« less

Authors:
 [1];  [1];  [1];  [1];  [2];  [2];  [2];  [2];  [2]; ORCiD logo [3];  [3];  [3]; ORCiD logo [3];  [3]
  1. LBL, Berkeley
  2. Caltech
  3. Fermilab
Publication Date:
Research Org.:
Lawrence Berkeley National Lab. (LBNL), Berkeley, CA (United States); Fermi National Accelerator Lab. (FNAL), Batavia, IL (United States)
Sponsoring Org.:
USDOE Office of Science (SC), High Energy Physics (HEP) (SC-25)
OSTI Identifier:
1484458
Report Number(s):
arXiv:1810.06111; FERMILAB-CONF-18-598-CD
1698413
DOE Contract Number:  
AC02-07CH11359
Resource Type:
Conference
Resource Relation:
Conference: 4th International Workshop Connecting The Dots 2018, Seattle, Washington, USA, 03/20-03/22/2018
Country of Publication:
United States
Language:
English
Subject:
72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS

Citation Formats

Farrell, Steven, Calafiura, Paolo, Mudigonda, Mayur, Prabhat, Prabhat,, Anderson, Dustin, Vlimant, Jean-Roch, Zheng, Stephan, Bendavid, Josh, Spiropulu, Maria, Cerati, Giuseppe, Gray, Lindsey, Kowalkowski, Jim, Spentzouris, Panagiotis, and Tsaris, Aristeidis. Novel deep learning methods for track reconstruction. United States: N. p., 2018. Web.
Farrell, Steven, Calafiura, Paolo, Mudigonda, Mayur, Prabhat, Prabhat,, Anderson, Dustin, Vlimant, Jean-Roch, Zheng, Stephan, Bendavid, Josh, Spiropulu, Maria, Cerati, Giuseppe, Gray, Lindsey, Kowalkowski, Jim, Spentzouris, Panagiotis, & Tsaris, Aristeidis. Novel deep learning methods for track reconstruction. United States.
Farrell, Steven, Calafiura, Paolo, Mudigonda, Mayur, Prabhat, Prabhat,, Anderson, Dustin, Vlimant, Jean-Roch, Zheng, Stephan, Bendavid, Josh, Spiropulu, Maria, Cerati, Giuseppe, Gray, Lindsey, Kowalkowski, Jim, Spentzouris, Panagiotis, and Tsaris, Aristeidis. Sun . "Novel deep learning methods for track reconstruction". United States. https://www.osti.gov/servlets/purl/1484458.
@article{osti_1484458,
title = {Novel deep learning methods for track reconstruction},
author = {Farrell, Steven and Calafiura, Paolo and Mudigonda, Mayur and Prabhat, Prabhat, and Anderson, Dustin and Vlimant, Jean-Roch and Zheng, Stephan and Bendavid, Josh and Spiropulu, Maria and Cerati, Giuseppe and Gray, Lindsey and Kowalkowski, Jim and Spentzouris, Panagiotis and Tsaris, Aristeidis},
abstractNote = {For the past year, the HEP.TrkX 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.},
doi = {},
journal = {},
number = ,
volume = ,
place = {United States},
year = {2018},
month = {10}
}

Conference:
Other availability
Please see Document Availability for additional information on obtaining the full-text document. Library patrons may search WorldCat to identify libraries that hold this conference proceeding.

Save / Share: