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

Title: Weakly supervised classification in high energy physics

Abstract

As machine learning algorithms become increasingly sophisticated to exploit subtle features of the data, they often become more dependent on simulations. Here, this paper presents a new approach called weakly supervised classification in which class proportions are the only input into the machine learning algorithm. Using one of the most challenging binary classification tasks in high energy physics $-$ quark versus gluon tagging $-$ we show that weakly supervised classification can match the performance of fully supervised algorithms. Furthermore, by design, the new algorithm is insensitive to any mis-modeling of discriminating features in the data by the simulation. Weakly supervised classification is a general procedure that can be applied to a wide variety of learning problems to boost performance and robustness when detailed simulations are not reliable or not available.

Authors:
 [1];  [2]; ORCiD logo [3];  [3]
  1. Stanford Univ., CA (United States). Dept. of Physics
  2. Lawrence Berkeley National Lab. (LBNL), Berkeley, CA (United States). Physics Division
  3. SLAC National Accelerator Lab., Menlo Park, CA (United States)
Publication Date:
Research Org.:
SLAC National Accelerator Lab., Menlo Park, CA (United States); Lawrence Berkeley National Lab. (LBNL), Berkeley, CA (United States)
Sponsoring Org.:
USDOE
OSTI Identifier:
1369428
Alternate Identifier(s):
OSTI ID: 1379863
Grant/Contract Number:
AC02-76SF00515; AC02-05CH11231
Resource Type:
Journal Article: Accepted Manuscript
Journal Name:
Journal of High Energy Physics (Online)
Additional Journal Information:
Journal Name: Journal of High Energy Physics (Online); Journal Volume: 2017; Journal Issue: 5; Journal ID: ISSN 1029-8479
Publisher:
Springer Berlin
Country of Publication:
United States
Language:
English
Subject:
72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS; 97 MATHEMATICS AND COMPUTING; Jets

Citation Formats

Dery, Lucio Mwinmaarong, Nachman, Benjamin, Rubbo, Francesco, and Schwartzman, Ariel. Weakly supervised classification in high energy physics. United States: N. p., 2017. Web. doi:10.1007/JHEP05(2017)145.
Dery, Lucio Mwinmaarong, Nachman, Benjamin, Rubbo, Francesco, & Schwartzman, Ariel. Weakly supervised classification in high energy physics. United States. doi:10.1007/JHEP05(2017)145.
Dery, Lucio Mwinmaarong, Nachman, Benjamin, Rubbo, Francesco, and Schwartzman, Ariel. Mon . "Weakly supervised classification in high energy physics". United States. doi:10.1007/JHEP05(2017)145. https://www.osti.gov/servlets/purl/1369428.
@article{osti_1369428,
title = {Weakly supervised classification in high energy physics},
author = {Dery, Lucio Mwinmaarong and Nachman, Benjamin and Rubbo, Francesco and Schwartzman, Ariel},
abstractNote = {As machine learning algorithms become increasingly sophisticated to exploit subtle features of the data, they often become more dependent on simulations. Here, this paper presents a new approach called weakly supervised classification in which class proportions are the only input into the machine learning algorithm. Using one of the most challenging binary classification tasks in high energy physics $-$ quark versus gluon tagging $-$ we show that weakly supervised classification can match the performance of fully supervised algorithms. Furthermore, by design, the new algorithm is insensitive to any mis-modeling of discriminating features in the data by the simulation. Weakly supervised classification is a general procedure that can be applied to a wide variety of learning problems to boost performance and robustness when detailed simulations are not reliable or not available.},
doi = {10.1007/JHEP05(2017)145},
journal = {Journal of High Energy Physics (Online)},
number = 5,
volume = 2017,
place = {United States},
year = {Mon May 01 00:00:00 EDT 2017},
month = {Mon May 01 00:00:00 EDT 2017}
}

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

Citation Metrics:
Cited by: 3works
Citation information provided by
Web of Science

Save / Share:
  • The investigation and classification of neutron energy spectra is one of the most complicated and important research problems. In order to eliminate the influence of possible systematic errors upon the classification, one must make a representative sampling of data which were measured with a single technique within particular energy limits. For this purpose, on the 70-GeV proton accelerator of the Institute of High-Energy Physics, there were measured more than 100 spectra behind the shield at points which can be characterized by variations of well distinguishable parameters such as thickness and composition of the biological shield, inhomogeneities (passageways and labyrinths) inmore » the shield, the source-shield-detector geometry, and the specific conditions of accelerator operation.« less
  • Cited by 12
  • Modern machine learning techniques can be used to construct powerful models for difficult collider physics problems. In many applications, however, these models are trained on imperfect simulations due to a lack of truth-level information in the data, which risks the model learning artifacts of the simulation. In this paper, we introduce the paradigm of classification without labels (CWoLa) in which a classifier is trained to distinguish statistical mixtures of classes, which are common in collider physics. Crucially, neither individual labels nor class proportions are required, yet we prove that the optimal classifier in the CWoLa paradigm is also the optimalmore » classifier in the traditional fully-supervised case where all label information is available. After demonstrating the power of this method in an analytical toy example, we consider a realistic benchmark for collider physics: distinguishing quark- versus gluon-initiated jets using mixed quark/gluon training samples. More generally, CWoLa can be applied to any classification problem where labels or class proportions are unknown or simulations are unreliable, but statistical mixtures of the classes are available.« less