Deep-learning jets with uncertainties and more

Deep-learning jets with uncertainties and more

Bayesian neural networks allow us to keep track of uncertainties, for example in top tagging, by learning a tagger output together with an error band. We illustrate the main features of Bayesian versions of established deep-learning taggers. We show how they capture statistical uncertainties from...

Full description

Bibliographic Details
Main Author: Sven Bollweg, Manuel Haussmann, Gregor Kasieczka, Michel Luchmann, Tilman Plehn, Jennifer Thompson
Format: Article
Language:English
Published: SciPost 2020-01-01
Series:SciPost Physics
Online Access:https://scipost.org/SciPostPhys.8.1.006
id doaj-ec6210b12d8546ecb84f625917d92a5e
record_format Article
spelling doaj-ec6210b12d8546ecb84f625917d92a5e2020-11-25T02:25:49ZengSciPostSciPost Physics2542-46532020-01-018100610.21468/SciPostPhys.8.1.006Deep-learning jets with uncertainties and moreSven Bollweg, Manuel Haussmann, Gregor Kasieczka, Michel Luchmann, Tilman Plehn, Jennifer ThompsonBayesian neural networks allow us to keep track of uncertainties, for example in top tagging, by learning a tagger output together with an error band. We illustrate the main features of Bayesian versions of established deep-learning taggers. We show how they capture statistical uncertainties from finite training samples, systematics related to the jet energy scale, and stability issues through pile-up. Altogether, Bayesian networks offer many new handles to understand and control deep learning at the LHC without introducing a visible prior effect and without compromising the network performance.https://scipost.org/SciPostPhys.8.1.006
collection DOAJ
language English
format Article
sources DOAJ
author Sven Bollweg, Manuel Haussmann, Gregor Kasieczka, Michel Luchmann, Tilman Plehn, Jennifer Thompson
spellingShingle Sven Bollweg, Manuel Haussmann, Gregor Kasieczka, Michel Luchmann, Tilman Plehn, Jennifer Thompson
Deep-learning jets with uncertainties and more
SciPost Physics
author_facet Sven Bollweg, Manuel Haussmann, Gregor Kasieczka, Michel Luchmann, Tilman Plehn, Jennifer Thompson
author_sort Sven Bollweg, Manuel Haussmann, Gregor Kasieczka, Michel Luchmann, Tilman Plehn, Jennifer Thompson
title Deep-learning jets with uncertainties and more
title_short Deep-learning jets with uncertainties and more
title_full Deep-learning jets with uncertainties and more
title_fullStr Deep-learning jets with uncertainties and more
title_full_unstemmed Deep-learning jets with uncertainties and more
title_sort deep-learning jets with uncertainties and more
publisher SciPost
series SciPost Physics
issn 2542-4653
publishDate 2020-01-01
description Bayesian neural networks allow us to keep track of uncertainties, for example in top tagging, by learning a tagger output together with an error band. We illustrate the main features of Bayesian versions of established deep-learning taggers. We show how they capture statistical uncertainties from finite training samples, systematics related to the jet energy scale, and stability issues through pile-up. Altogether, Bayesian networks offer many new handles to understand and control deep learning at the LHC without introducing a visible prior effect and without compromising the network performance.
url https://scipost.org/SciPostPhys.8.1.006
work_keys_str_mv AT svenbollwegmanuelhaussmanngregorkasieczkamichelluchmanntilmanplehnjenniferthompson deeplearningjetswithuncertaintiesandmore
_version_ 1724850097619992576

Similar Items