Scalable deep convolutional neural networks for sparse, locally dense liquid argon time projection chamber data
Abstract
Deep convolutional neural networks (CNNs) show strong promise for analyzing scientific data in many domains including particle imaging detectors such as a liquid argon time projection chamber (LArTPC). Yet the high sparsity of LArTPC data challenges traditional CNNs which were designed for dense data such as photographs. A naive application of CNNs on LArTPC data results in inefficient computations and a poor scalability to large LArTPC detectors such as the Short Baseline Neutrino Program and Deep Underground Neutrino Experiment. Recently Submanifold Sparse Convolutional Networks (SSCNs) have been proposed to address this class of challenges. We report their performance on a 3D semantic segmentation task on simulated LArTPC samples. In comparison with standard CNNs, we observe that the computation memory and wall-time cost for inference are reduced by a factor of 364 and 33, respectively, without loss of accuracy. The same factors for 2D samples are found to be 93 and 3.1 respectively. Using SSCN and public 3D LArTPC samples, we present the first machine learning-based approach to the reconstruction of Michel electrons, a standard candle for energy calibration in LArTPC due to their very well understood energy spectrum. We find a Michel electrons identification efficiency of 93.9% and a 96.7%more »
- Authors:
- Publication Date:
- Research Org.:
- SLAC National Accelerator Laboratory (SLAC), Menlo Park, CA (United States)
- Sponsoring Org.:
- USDOE Office of Science (SC)
- OSTI Identifier:
- 1638112
- Alternate Identifier(s):
- OSTI ID: 1646801
- Grant/Contract Number:
- AC02-76SF00515
- Resource Type:
- Published Article
- Journal Name:
- Physical Review. D.
- Additional Journal Information:
- Journal Name: Physical Review. D. Journal Volume: 102 Journal Issue: 1; Journal ID: ISSN 2470-0010
- Publisher:
- American Physical Society (APS)
- Country of Publication:
- United States
- Language:
- English
- Subject:
- 72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS
Citation Formats
Dominé, Laura, Terao, Kazuhiro, and DeepLearnPhysics Collaboration. Scalable deep convolutional neural networks for sparse, locally dense liquid argon time projection chamber data. United States: N. p., 2020.
Web. doi:10.1103/PhysRevD.102.012005.
Dominé, Laura, Terao, Kazuhiro, & DeepLearnPhysics Collaboration. Scalable deep convolutional neural networks for sparse, locally dense liquid argon time projection chamber data. United States. https://doi.org/10.1103/PhysRevD.102.012005
Dominé, Laura, Terao, Kazuhiro, and DeepLearnPhysics Collaboration. Fri .
"Scalable deep convolutional neural networks for sparse, locally dense liquid argon time projection chamber data". United States. https://doi.org/10.1103/PhysRevD.102.012005.
@article{osti_1638112,
title = {Scalable deep convolutional neural networks for sparse, locally dense liquid argon time projection chamber data},
author = {Dominé, Laura and Terao, Kazuhiro and DeepLearnPhysics Collaboration},
abstractNote = {Deep convolutional neural networks (CNNs) show strong promise for analyzing scientific data in many domains including particle imaging detectors such as a liquid argon time projection chamber (LArTPC). Yet the high sparsity of LArTPC data challenges traditional CNNs which were designed for dense data such as photographs. A naive application of CNNs on LArTPC data results in inefficient computations and a poor scalability to large LArTPC detectors such as the Short Baseline Neutrino Program and Deep Underground Neutrino Experiment. Recently Submanifold Sparse Convolutional Networks (SSCNs) have been proposed to address this class of challenges. We report their performance on a 3D semantic segmentation task on simulated LArTPC samples. In comparison with standard CNNs, we observe that the computation memory and wall-time cost for inference are reduced by a factor of 364 and 33, respectively, without loss of accuracy. The same factors for 2D samples are found to be 93 and 3.1 respectively. Using SSCN and public 3D LArTPC samples, we present the first machine learning-based approach to the reconstruction of Michel electrons, a standard candle for energy calibration in LArTPC due to their very well understood energy spectrum. We find a Michel electrons identification efficiency of 93.9% and a 96.7% purity. Reconstructed Michel electron clusters yield 95.4% in average pixel clustering efficiency and 95.5% in purity. The results are compelling in showing the strong promise of scalable data reconstruction technique using deep neural networks for large scale LArTPC detectors.},
doi = {10.1103/PhysRevD.102.012005},
journal = {Physical Review. D.},
number = 1,
volume = 102,
place = {United States},
year = {Fri Jul 10 00:00:00 EDT 2020},
month = {Fri Jul 10 00:00:00 EDT 2020}
}
https://doi.org/10.1103/PhysRevD.102.012005
Web of Science
Works referenced in this record:
Geant4—a simulation toolkit
journal, July 2003
- Agostinelli, S.; Allison, J.; Amako, K.
- Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, Vol. 506, Issue 3
Convolutional neural networks applied to neutrino events in a liquid argon time projection chamber
journal, March 2017
- Acciarri, R.; Adams, C.; An, R.
- Journal of Instrumentation, Vol. 12, Issue 03
Machine learning at the energy and intensity frontiers of particle physics
journal, August 2018
- Radovic, Alexander; Williams, Mike; Rousseau, David
- Nature, Vol. 560, Issue 7716
Michel electron reconstruction using cosmic-ray data from the MicroBooNE LArTPC
journal, September 2017
- Acciarri, R.; Adams, C.; An, R.
- Journal of Instrumentation, Vol. 12, Issue 09
Deep learning
journal, May 2015
- LeCun, Yann; Bengio, Yoshua; Hinton, Geoffrey
- Nature, Vol. 521, Issue 7553
LArSoft: toolkit for simulation, reconstruction and analysis of liquid argon TPC neutrino detectors
journal, October 2017
- Snider, E. L.; Petrillo, G.
- Journal of Physics: Conference Series, Vol. 898
Enhancing reproducibility in scientific computing: Metrics and registry for Singularity containers
journal, November 2017
- Sochat, Vanessa V.; Prybol, Cameron J.; Kurtzer, Gregory M.
- PLOS ONE, Vol. 12, Issue 11
Design and construction of the MicroBooNE detector
journal, February 2017
- Acciarri, R.; Adams, C.; An, R.
- Journal of Instrumentation, Vol. 12, Issue 02
Measurement of the μ decay spectrum with the ICARUS liquid Argon TPC
journal, March 2004
- The ICARUS Collaboration,
- The European Physical Journal C, Vol. 33, Issue 2, p. 233-241
Design, construction and tests of the ICARUS T600 detector
journal, July 2004
- Amerio, S.; Amoruso, S.; Antonello, M.
- Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, Vol. 527, Issue 3
Deep neural network for pixel-level electromagnetic particle identification in the MicroBooNE liquid argon time projection chamber
journal, May 2019
- Adams, C.; Alrashed, M.; An, R.
- Physical Review D, Vol. 99, Issue 9