Understanding Event-Generation Networks via Uncertainties

Bellagente, Marco; Haussmann, Manuel; Luchmann, Michel; Plehn, Tilman

doi:10.21468/SciPostPhys.13.1.003

SciPost Physics

Understanding Event-Generation Networks via Uncertainties

Marco Bellagente, Manuel Haußmann, Michel Luchmann, Tilman Plehn

SciPost Phys. 13, 003 (2022) · published 21 July 2022

doi: 10.21468/SciPostPhys.13.1.003
pdf
Submissions/Reports

Abstract

Following the growing success of generative neural networks in LHC simulations, the crucial question is how to control the networks and assign uncertainties to their event output. We show how Bayesian normalizing flow or invertible networks capture uncertainties from the training and turn them into an uncertainty on the event weight. Fundamentally, the interplay between density and uncertainty estimates indicates that these networks learn functions in analogy to parameter fits rather than binned event counts.

TY  - JOUR
PB  - SciPost Foundation
DO  - 10.21468/SciPostPhys.13.1.003
TI  - Understanding Event-Generation Networks via Uncertainties
PY  - 2022/07/21
UR  - https://www.scipost.org/SciPostPhys.13.1.003
JF  - SciPost Physics
JA  - SciPost Phys.
VL  - 13
IS  - 1
SP  - 003
A1  - Bellagente, Marco
AU  - Haussmann, Manuel
AU  - Luchmann, Michel
AU  - Plehn, Tilman
AB  - Following the growing success of generative neural networks in LHC simulations, the crucial question is how to control the networks and assign uncertainties to their event output. We show how Bayesian normalizing flow or invertible networks capture uncertainties from the training and turn them into an uncertainty on the event weight. Fundamentally, the interplay between density and uncertainty estimates indicates that these networks learn functions in analogy to parameter fits rather than binned event counts.
ER  -

@Article{10.21468/SciPostPhys.13.1.003,
	title={{Understanding Event-Generation Networks via Uncertainties}},
	author={Marco Bellagente and Manuel Haußmann and Michel Luchmann and Tilman Plehn},
	journal={SciPost Phys.},
	volume={13},
	pages={003},
	year={2022},
	publisher={SciPost},
	doi={10.21468/SciPostPhys.13.1.003},
	url={https://scipost.org/10.21468/SciPostPhys.13.1.003},
}

Cited by 26

Campbell et al., Event generators for high-energy physics experiments
SciPost Phys. 16, 130 (2024) [Crossref]
Bieringer et al., Calomplification — the power of generative calorimeter models
J. Inst. 17, P09028 (2022) [Crossref]
Khosa et al., A simple guide from machine learning outputs to statistical criteria in particle physics
SciPost Phys. Core 5, 050 (2022) [Crossref]
Konar et al., Hypergraphs in LHC phenomenology — the next frontier of IRC-safe feature extraction
J. High Energ. Phys. 2024, 113 (2024) [Crossref]
Barman et al., Top-philic machine learning
Eur. Phys. J. Spec. Top., (2024) [Crossref]
Krause et al., Fast and accurate simulations of calorimeter showers with normalizing flows
Phys. Rev. D 107, 113003 (2023) [Crossref]
Verheyen, Event Generation and Density Estimation with Surjective Normalizing Flows
SciPost Phys. 13, 047 (2022) [Crossref]
Butter et al., Generative networks for precision enthusiasts
SciPost Phys. 14, 078 (2023) [Crossref]
Butter et al., Performance versus resilience in modern quark-gluon tagging
SciPost Phys. Core 6, 085 (2023) [Crossref]
Das et al., How to understand limitations of generative networks
SciPost Phys. 16, 031 (2024) [Crossref]
Krause et al., CaloFlow for CaloChallenge dataset 1
SciPost Phys. 16, 126 (2024) [Crossref]
Krause et al., Anomaly detection with flow-based fast calorimeter simulators
Phys. Rev. D 110, 035036 (2024) [Crossref]
Gavranovič et al., Systematic evaluation of generative machine learning capability to simulate distributions of observables at the large hadron collider
Eur. Phys. J. C 84, 911 (2024) [Crossref]
Buss et al., What's anomalous in LHC jets?
SciPost Phys. 15, 168 (2023) [Crossref]
Deutschmann et al., Accelerating HEP simulations with Neural Importance Sampling
J. High Energ. Phys. 2024, 83 (2024) [Crossref]
Ackerschott et al., Returning CP-observables to the frames they belong
SciPost Phys. 17, 001 (2024) [Crossref]
Butter et al., Machine learning and LHC event generation
SciPost Phys. 14, 079 (2023) [Crossref]
Badger et al., Loop amplitudes from precision networks
SciPost Phys. Core 6, 034 (2023) [Crossref]
Karagiorgi et al., Machine learning in the search for new fundamental physics
Nat Rev Phys 4, 399 (2022) [Crossref]
Heimel et al., MadNIS - Neural multi-channel importance sampling
SciPost Phys. 15, 141 (2023) [Crossref]
Butter et al., Ephemeral Learning - Augmenting Triggers with Online-Trained Normalizing Flows
SciPost Phys. 13, 087 (2022) [Crossref]
Butter et al., Kicking it off(-shell) with direct diffusion
SciPost Phys. Core 7, 064 (2024) [Crossref]
Yallup et al., Exploring phase space with nested sampling
Eur. Phys. J. C 82, 678 (2022) [Crossref]
Alvarez et al., Exploring unsupervised top tagging using Bayesian inference
SciPost Phys. Core 6, 046 (2023) [Crossref]
Goodsell et al., Active learning BSM parameter spaces
Eur. Phys. J. C 83, 268 (2023) [Crossref]
Butter et al., Two invertible networks for the matrix element method
SciPost Phys. 15, 094 (2023) [Crossref]

Authors / Affiliation: mappings to Contributors and Organizations

See all Organizations.

¹ Ruprecht-Karls-Universität Heidelberg / Heidelberg University

Funder for the research work leading to this publication

Deutsche Forschungsgemeinschaft / German Research FoundationDeutsche Forschungsgemeinschaft [DFG]