https://doi.org/10.1140/epjc/s10052-022-10944-3
Regular Article - Theoretical Physics
A method for approximating optimal statistical significances with machine-learned likelihoods
1
Instituto de Física Teórica UAM-CSIC, C/ Nicolás Cabrera 13-15, Campus de Cantoblanco, 28049, Madrid, Spain
2
IFLP, CONICET-Dpto. de Física, Universidad Nacional de La Plata, C.C. 67, 1900, La Plata, Argentina
3
Departamento de Física Teórica, Universidad Autónoma de Madrid, 28049, Cantoblanco, Madrid, Spain
4
Departament de Física Teòrica and IFIC, Universitat de València-CSIC, 46100, Burjassot, Spain
5
Deutsches Elektronen-Synchrotron DESY, Notkestr. 85, 22607, Hamburg, Germany
6
Jožef Stefan Institute, Jamova 39, 1000, Ljubljana, Slovenia
7
International Center for Advanced Studies (ICAS) and ICIFI, UNSAM, Campus Miguelete, 25 de Mayo y Francia, CP1650, San Martín, Buenos Aires, Argentina
Received:
18
May
2022
Accepted:
23
October
2022
Published online:
5
November
2022
Machine-learning techniques have become fundamental in high-energy physics and, for new physics searches, it is crucial to know their performance in terms of experimental sensitivity, understood as the statistical significance of the signal-plus-background hypothesis over the background-only one. We present here a simple method that combines the power of current machine-learning techniques to face high-dimensional data with the likelihood-based inference tests used in traditional analyses, which allows us to estimate the sensitivity for both discovery and exclusion limits through a single parameter of interest, the signal strength. Based on supervised learning techniques, it can perform well also with high-dimensional data, when traditional techniques cannot. We apply the method to a toy model first, so we can explore its potential, and then to a LHC study of new physics particles in dijet final states. Considering as the optimal statistical significance the one we would obtain if the true generative functions were known, we show that our method provides a better approximation than the usual naive counting experimental results.
© The Author(s) 2022
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Funded by SCOAP3. SCOAP3 supports the goals of the International Year of Basic Sciences for Sustainable Development.