Comparing weak- and unsupervised methods for resonant anomaly detection
Abstract
Anomaly detection techniques are growing in importance at the Large Hadron Collider (LHC), motivated by the increasing need to search for new physics in a model-agnostic way. In this work, we provide a detailed comparative study between a well-studied unsupervised method called the autoencoder (AE) and a weakly-supervised approach based on the Classification Without Labels (CWoLa) technique. We examine the ability of the two methods to identify a new physics signal at different cross sections in a fully hadronic resonance search. By construction, the AE classification performance is independent of the amount of injected signal. In contrast, the CWoLa performance improves with increasing signal abundance. When integrating these approaches with a complete background estimate, we find that the two methods have complementary sensitivity. In particular, CWoLa is effective at finding diverse and moderately rare signals while the AE can provide sensitivity to very rare signals, but only with certain topologies. We therefore demonstrate that both techniques are complementary and can be used together for anomaly detection at the LHC.
- Authors:
- Publication Date:
- Research Org.:
- SLAC National Accelerator Laboratory (SLAC), Menlo Park, CA (United States); Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)
- Sponsoring Org.:
- USDOE Office of Science (SC), Basic Energy Sciences (BES); USDOE Office of Science (SC), High Energy Physics (HEP)
- OSTI Identifier:
- 1808209
- Alternate Identifier(s):
- OSTI ID: 1819743; OSTI ID: 1820604
- Grant/Contract Number:
- AC02-05CH11231; AC02-76SF00515; DOE-SC0010008
- Resource Type:
- Published Article
- Journal Name:
- European Physical Journal. C, Particles and Fields
- Additional Journal Information:
- Journal Name: European Physical Journal. C, Particles and Fields Journal Volume: 81 Journal Issue: 7; Journal ID: ISSN 1434-6044
- Publisher:
- Springer Science + Business Media
- Country of Publication:
- Germany
- Language:
- English
- Subject:
- 72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS
Citation Formats
Collins, Jack H., Martín-Ramiro, Pablo, Nachman, Benjamin, and Shih, David. Comparing weak- and unsupervised methods for resonant anomaly detection. Germany: N. p., 2021.
Web. doi:10.1140/epjc/s10052-021-09389-x.
Collins, Jack H., Martín-Ramiro, Pablo, Nachman, Benjamin, & Shih, David. Comparing weak- and unsupervised methods for resonant anomaly detection. Germany. https://doi.org/10.1140/epjc/s10052-021-09389-x
Collins, Jack H., Martín-Ramiro, Pablo, Nachman, Benjamin, and Shih, David. Thu .
"Comparing weak- and unsupervised methods for resonant anomaly detection". Germany. https://doi.org/10.1140/epjc/s10052-021-09389-x.
@article{osti_1808209,
title = {Comparing weak- and unsupervised methods for resonant anomaly detection},
author = {Collins, Jack H. and Martín-Ramiro, Pablo and Nachman, Benjamin and Shih, David},
abstractNote = {Anomaly detection techniques are growing in importance at the Large Hadron Collider (LHC), motivated by the increasing need to search for new physics in a model-agnostic way. In this work, we provide a detailed comparative study between a well-studied unsupervised method called the autoencoder (AE) and a weakly-supervised approach based on the Classification Without Labels (CWoLa) technique. We examine the ability of the two methods to identify a new physics signal at different cross sections in a fully hadronic resonance search. By construction, the AE classification performance is independent of the amount of injected signal. In contrast, the CWoLa performance improves with increasing signal abundance. When integrating these approaches with a complete background estimate, we find that the two methods have complementary sensitivity. In particular, CWoLa is effective at finding diverse and moderately rare signals while the AE can provide sensitivity to very rare signals, but only with certain topologies. We therefore demonstrate that both techniques are complementary and can be used together for anomaly detection at the LHC.},
doi = {10.1140/epjc/s10052-021-09389-x},
journal = {European Physical Journal. C, Particles and Fields},
number = 7,
volume = 81,
place = {Germany},
year = {Thu Jul 15 00:00:00 EDT 2021},
month = {Thu Jul 15 00:00:00 EDT 2021}
}
https://doi.org/10.1140/epjc/s10052-021-09389-x
Works referenced in this record:
QCD or what?
journal, January 2019
- Heimel, Theo; Kasieczka, Gregor; Plehn, Tilman
- SciPost Physics, Vol. 6, Issue 3
The anti- k t jet clustering algorithm
journal, April 2008
- Cacciari, Matteo; Salam, Gavin P.; Soyez, Gregory
- Journal of High Energy Physics, Vol. 2008, Issue 04
Quasi-model-independent search for new physics at large transverse momentum
journal, June 2001
- Abazov, V. M.; Abbott, B.; Abdesselam, A.
- Physical Review D, Vol. 64, Issue 1
Model-independent and quasi-model-independent search for new physics at CDF
journal, July 2008
- Aaltonen, T.; Abulencia, A.; Adelman, J.
- Physical Review D, Vol. 78, Issue 1
Maximizing boosted top identification by minimizing N-subjettiness
journal, February 2012
- Thaler, Jesse; Van Tilburg, Ken
- Journal of High Energy Physics, Vol. 2012, Issue 2
A brief introduction to PYTHIA 8.1
journal, June 2008
- Sjöstrand, Torbjörn; Mrenna, Stephen; Skands, Peter
- Computer Physics Communications, Vol. 178, Issue 11
Simulation assisted likelihood-free anomaly detection
journal, May 2020
- Andreassen, Anders; Nachman, Benjamin; Shih, David
- Physical Review D, Vol. 101, Issue 9
Learning new physics from a machine
journal, January 2019
- D’Agnolo, Raffaele Tito; Wulzer, Andrea
- Physical Review D, Vol. 99, Issue 1
Variational autoencoders for new physics mining at the Large Hadron Collider
journal, May 2019
- Cerri, Olmo; Nguyen, Thong Q.; Pierini, Maurizio
- Journal of High Energy Physics, Vol. 2019, Issue 5
DELPHES 3: a modular framework for fast simulation of a generic collider experiment
journal, February 2014
- de Favereau, J.; Delaere, C.; Demin, P.
- Journal of High Energy Physics, Vol. 2014, Issue 2
Adversarially-trained autoencoders for robust unsupervised new physics searches
journal, October 2019
- Blance, Andrew; Spannowsky, Michael; Waite, Philip
- Journal of High Energy Physics, Vol. 2019, Issue 10
Multivariate discrimination and the Higgs+W/Z search
journal, April 2011
- Gallicchio, Jason; Huth, John; Kagan, Michael
- Journal of High Energy Physics, Vol. 2011, Issue 4
FastJet user manual: (for version 3.0.2)
journal, March 2012
- Cacciari, Matteo; Salam, Gavin P.; Soyez, Gregory
- The European Physical Journal C, Vol. 72, Issue 3
Anomaly Detection for Resonant New Physics with Machine Learning
journal, December 2018
- Collins, Jack; Howe, Kiel; Nachman, Benjamin
- Physical Review Letters, Vol. 121, Issue 24
A generic anti-QCD jet tagger
journal, November 2017
- Aguilar-Saavedra, J. A.; Collins, Jack; Mishra, Rashmish K.
- Journal of High Energy Physics, Vol. 2017, Issue 11
A general search for new phenomena at HERA
journal, April 2009
- Aaron, F. D.; Alexa, C.; Andreev, V.
- Physics Letters B, Vol. 674, Issue 4-5
Uncovering latent jet substructure
journal, September 2019
- Dillon, Barry M.; Faroughy, Darius A.; Kamenik, Jernej F.
- Physical Review D, Vol. 100, Issue 5
Anomaly detection with density estimation
journal, April 2020
- Nachman, Benjamin; Shih, David
- Physical Review D, Vol. 101, Issue 7
Guiding new physics searches with unsupervised learning
journal, March 2019
- De Simone, Andrea; Jacques, Thomas
- The European Physical Journal C, Vol. 79, Issue 4
Global search for new physics with at CDF
journal, January 2009
- Aaltonen, T.; Adelman, J.; Akimoto, T.
- Physical Review D, Vol. 79, Issue 1
Identifying boosted objects with N-subjettiness
journal, March 2011
- Thaler, Jesse; Van Tilburg, Ken
- Journal of High Energy Physics, Vol. 2011, Issue 3
A general search for new phenomena in ep scattering at HERA
journal, November 2004
- Aktas, A.; Andreev, V.; Anthonis, T.
- Physics Letters B, Vol. 602, Issue 1-2
Quasi-Model-Independent Search for New High Physics at D0
journal, April 2001
- Abbott, B.; Abdesselam, A.; Abolins, M.
- Physical Review Letters, Vol. 86, Issue 17
Extending the search for new resonances with machine learning
journal, January 2019
- Collins, Jack H.; Howe, Kiel; Nachman, Benjamin
- Physical Review D, Vol. 99, Issue 1
Search for narrow and broad dijet resonances in proton-proton collisions at s = 13 $$ \sqrt{s}=13 $$ TeV and constraints on dark matter mediators and other new particles
journal, August 2018
- Sirunyan, A. M.; Tumasyan, A.; Adam, W.
- Journal of High Energy Physics, Vol. 2018, Issue 8
Transferability of deep learning models in searches for new physics at colliders
journal, February 2020
- Romão, M. Crispim; Castro, N. F.; Pedro, R.
- Physical Review D, Vol. 101, Issue 3
A strategy for a general search for new phenomena using data-derived signal regions and its application within the ATLAS experiment
journal, February 2019
- Aaboud, M.; Aad, G.; Abbott, B.
- The European Physical Journal C, Vol. 79, Issue 2
Search for new physics in data at DØ using SLEUTH: A quasi-model-independent search strategy for new physics
journal, October 2000
- Abbott, B.; Abolins, M.; Abramov, V.
- Physical Review D, Vol. 62, Issue 9