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

Title: Reducing model bias in a deep learning classifier using domain adversarial neural networks in the MINERvA experiment

Abstract

We present a simulation-based study using deep convolutional neural networks (DCNNs) to identify neutrino interaction vertices in the MINERvA passive targets region, and illustrate the application of domain adversarial neural networks (DANNs) in this context. DANNs are designed to be trained in one domain (simulated data) but tested in a second domain (physics data) and utilize unlabeled data from the second domain so that during training only features which are unable to discriminate between the domains are promoted. MINERvA is a neutrino-nucleus scattering experiment using the NuMI beamline at Fermilab. A-dependent cross sections are an important part of the physics program, and these measurements require vertex finding in complicated events. To illustrate the impact of the DANN we used a modified set of simulation in place of physics data during the training of the DANN and then used the label of the modified simulation during the evaluation of the DANN. In conclusion, we find that deep learning based methods offer significant advantages over our prior track-based reconstruction for the task of vertex finding, and that DANNs are able to improve the performance of deep networks by leveraging available unlabeled data and by mitigating network performance degradation rooted in biases inmore » the physics models used for training.« less

Authors:
; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; more »; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; « less
Publication Date:
Research Org.:
Oak Ridge National Laboratory, Oak Ridge Leadership Computing Facility (OLCF); Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States); Fermi National Accelerator Lab. (FNAL), Batavia, IL (United States)
Sponsoring Org.:
National Science Foundation (NSF); CNPq (Brazil); CoNaCyT (Mexico); IDI/IGI-UNI (Peru); Latin American Center for Physics (CLAF); Russian Ministry of Education and Science (Russia); National Science Centre of Poland; USDOE Office of Science (SC), High Energy Physics (HEP) (SC-25)
Contributing Org.:
MINERvA collaboration; MINERvA Collaboration
OSTI Identifier:
1487058
Alternate Identifier(s):
OSTI ID: 1484978
Report Number(s):
arXiv:1808.08332; FERMILAB-PUB-18-432-CD-ND
Journal ID: ISSN 1748-0221
Grant/Contract Number:  
AC05-00OR22725; AC02-07CH11359
Resource Type:
Accepted Manuscript
Journal Name:
Journal of Instrumentation
Additional Journal Information:
Journal Volume: 13; Journal Issue: 11; Journal ID: ISSN 1748-0221
Publisher:
Institute of Physics (IOP)
Country of Publication:
United States
Language:
English
Subject:
72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS; 97 MATHEMATICS AND COMPUTING; Analysis and statistical methods; Pattern recognition, cluster finding, calibration and fitting methods; Neutrino detectors

Citation Formats

Perdue, G. N., Ghosh, A., Wospakrik, M., Akbar, F., Andrade, D. A., Ascencio, M., Bellantoni, L., Bercellie, A., Betancourt, M., Vera, G. F. R. Caceres, Cai, T., Carneiro, M. F., Chaves, J., Coplowe, D., Motta, H. da, Díaz, G. A., Felix, J., Fields, L., Fine, R., Gago, A. M., Galindo, R., Golan, T., Gran, R., Han, J. Y., Harris, D. A., Jena, D., Kleykamp, J., Kordosky, M., Lu, X. -G., Maher, E., Mann, W. A., Marshall, C. M., McFarland, K. S., McGowan, A. M., Messerly, B., Miller, J., Nelson, J. K., Nguyen, C., Norrick, A., Nuruzzaman, Nuruzzaman, Olivier, A., Patton, R., Ramírez, M. A., Ransome, R. D., Ray, H., Ren, L., Rimal, D., Ruterbories, D., Schellman, H., Salinas, C. J. Solano, Su, H., Upadhyay, S., Valencia, E., Wolcott, J., Yaeggy, B., and Young, S. Reducing model bias in a deep learning classifier using domain adversarial neural networks in the MINERvA experiment. United States: N. p., 2018. Web. doi:10.1088/1748-0221/13/11/P11020.
Perdue, G. N., Ghosh, A., Wospakrik, M., Akbar, F., Andrade, D. A., Ascencio, M., Bellantoni, L., Bercellie, A., Betancourt, M., Vera, G. F. R. Caceres, Cai, T., Carneiro, M. F., Chaves, J., Coplowe, D., Motta, H. da, Díaz, G. A., Felix, J., Fields, L., Fine, R., Gago, A. M., Galindo, R., Golan, T., Gran, R., Han, J. Y., Harris, D. A., Jena, D., Kleykamp, J., Kordosky, M., Lu, X. -G., Maher, E., Mann, W. A., Marshall, C. M., McFarland, K. S., McGowan, A. M., Messerly, B., Miller, J., Nelson, J. K., Nguyen, C., Norrick, A., Nuruzzaman, Nuruzzaman, Olivier, A., Patton, R., Ramírez, M. A., Ransome, R. D., Ray, H., Ren, L., Rimal, D., Ruterbories, D., Schellman, H., Salinas, C. J. Solano, Su, H., Upadhyay, S., Valencia, E., Wolcott, J., Yaeggy, B., & Young, S. Reducing model bias in a deep learning classifier using domain adversarial neural networks in the MINERvA experiment. United States. doi:10.1088/1748-0221/13/11/P11020.
Perdue, G. N., Ghosh, A., Wospakrik, M., Akbar, F., Andrade, D. A., Ascencio, M., Bellantoni, L., Bercellie, A., Betancourt, M., Vera, G. F. R. Caceres, Cai, T., Carneiro, M. F., Chaves, J., Coplowe, D., Motta, H. da, Díaz, G. A., Felix, J., Fields, L., Fine, R., Gago, A. M., Galindo, R., Golan, T., Gran, R., Han, J. Y., Harris, D. A., Jena, D., Kleykamp, J., Kordosky, M., Lu, X. -G., Maher, E., Mann, W. A., Marshall, C. M., McFarland, K. S., McGowan, A. M., Messerly, B., Miller, J., Nelson, J. K., Nguyen, C., Norrick, A., Nuruzzaman, Nuruzzaman, Olivier, A., Patton, R., Ramírez, M. A., Ransome, R. D., Ray, H., Ren, L., Rimal, D., Ruterbories, D., Schellman, H., Salinas, C. J. Solano, Su, H., Upadhyay, S., Valencia, E., Wolcott, J., Yaeggy, B., and Young, S. Mon . "Reducing model bias in a deep learning classifier using domain adversarial neural networks in the MINERvA experiment". United States. doi:10.1088/1748-0221/13/11/P11020. https://www.osti.gov/servlets/purl/1487058.
@article{osti_1487058,
title = {Reducing model bias in a deep learning classifier using domain adversarial neural networks in the MINERvA experiment},
author = {Perdue, G. N. and Ghosh, A. and Wospakrik, M. and Akbar, F. and Andrade, D. A. and Ascencio, M. and Bellantoni, L. and Bercellie, A. and Betancourt, M. and Vera, G. F. R. Caceres and Cai, T. and Carneiro, M. F. and Chaves, J. and Coplowe, D. and Motta, H. da and Díaz, G. A. and Felix, J. and Fields, L. and Fine, R. and Gago, A. M. and Galindo, R. and Golan, T. and Gran, R. and Han, J. Y. and Harris, D. A. and Jena, D. and Kleykamp, J. and Kordosky, M. and Lu, X. -G. and Maher, E. and Mann, W. A. and Marshall, C. M. and McFarland, K. S. and McGowan, A. M. and Messerly, B. and Miller, J. and Nelson, J. K. and Nguyen, C. and Norrick, A. and Nuruzzaman, Nuruzzaman and Olivier, A. and Patton, R. and Ramírez, M. A. and Ransome, R. D. and Ray, H. and Ren, L. and Rimal, D. and Ruterbories, D. and Schellman, H. and Salinas, C. J. Solano and Su, H. and Upadhyay, S. and Valencia, E. and Wolcott, J. and Yaeggy, B. and Young, S.},
abstractNote = {We present a simulation-based study using deep convolutional neural networks (DCNNs) to identify neutrino interaction vertices in the MINERvA passive targets region, and illustrate the application of domain adversarial neural networks (DANNs) in this context. DANNs are designed to be trained in one domain (simulated data) but tested in a second domain (physics data) and utilize unlabeled data from the second domain so that during training only features which are unable to discriminate between the domains are promoted. MINERvA is a neutrino-nucleus scattering experiment using the NuMI beamline at Fermilab. A-dependent cross sections are an important part of the physics program, and these measurements require vertex finding in complicated events. To illustrate the impact of the DANN we used a modified set of simulation in place of physics data during the training of the DANN and then used the label of the modified simulation during the evaluation of the DANN. In conclusion, we find that deep learning based methods offer significant advantages over our prior track-based reconstruction for the task of vertex finding, and that DANNs are able to improve the performance of deep networks by leveraging available unlabeled data and by mitigating network performance degradation rooted in biases in the physics models used for training.},
doi = {10.1088/1748-0221/13/11/P11020},
journal = {Journal of Instrumentation},
number = 11,
volume = 13,
place = {United States},
year = {2018},
month = {11}
}

Journal Article:
Free Publicly Available Full Text
Publisher's Version of Record

Citation Metrics:
Cited by: 4 works
Citation information provided by
Web of Science

Save / Share: