Deep neural networks to recover unknown physical parameters from oscillating time series

Publikationstyp:
Zeitschriftenaufsatz
Metadaten:
Autoren
  • Antoine Garcon
  • Julian Vexler
  • Dmitry Budker
  • Stefan Kramer
Autoren-URL
https://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=fis-test-1&SrcAuth=WosAPI&KeyUT=WOS:000834668000009&DestLinkType=FullRecord&DestApp=WOS_CPL
DOI
10.1371/journal.pone.0268439
Externe Identifier
  • Clarivate Analytics Document Solution ID: 3L3MJ
  • PubMed Identifier: 35560322
ISSN
1932-6203
Ausgabe der Veröffentlichung
5
Zeitschrift
PLOS ONE
Artikelnummer
ARTN e0268439
Datum der Veröffentlichung
2022
Status
Published
Titel
Deep neural networks to recover unknown physical parameters from oscillating time series
Sub types
  • Article
Ausgabe der Zeitschrift
17

Datenquelle: Web of Science (Lite)

Andere Metadatenquellen:
Abstract
<jats:p>Deep neural networks are widely used in pattern-recognition tasks for which a human-comprehensible, quantitative description of the data-generating process, cannot be obtained. While doing so, neural networks often produce an abstract (entangled and non-interpretable) representation of the data-generating process. This may be one of the reasons why neural networks are not yet used extensively in physics-experiment signal processing: physicists generally require their analyses to yield quantitative information about the system they study. In this article we use a deep neural network to disentangle components of oscillating time series. To this aim, we design and train the neural network on synthetic oscillating time series to perform two tasks: a<jats:italic>regression</jats:italic>of the signal latent parameters and<jats:italic>signal denoising</jats:italic>by an<jats:italic>Autoencoder</jats:italic>-like architecture. We show that the regression and denoising performance is similar to those of least-square curve fittings with true latent-parameters initial guesses, in spite of the neural network needing no initial guesses at all. We then explore various applications in which we believe our architecture could prove useful for time-series processing, when prior knowledge is incomplete. As an example, we employ the neural network as a preprocessing tool to inform the least-square fits when initial guesses are unknown. Moreover, we show that the regression can be performed on some latent parameters, while ignoring the existence of others. Because the<jats:italic>Autoencoder</jats:italic>needs no prior information about the physical model, the remaining unknown latent parameters can still be captured, thus making use of partial prior knowledge, while leaving space for data exploration and discoveries.</jats:p>
Autoren
  • Antoine Garcon
  • Julian Vexler
  • Dmitry Budker
  • Stefan Kramer
DOI
10.1371/journal.pone.0268439
Editoren
  • Sheetal Kalyani
eISSN
1932-6203
Ausgabe der Veröffentlichung
5
Zeitschrift
PLOS ONE
Sprache
en
Online publication date
2022
Paginierung
e0268439 - e0268439
Status
Published online
Herausgeber
Public Library of Science (PLoS)
Herausgeber URL
http://dx.doi.org/10.1371/journal.pone.0268439
Datum der Datenerfassung
2023
Titel
Deep neural networks to recover unknown physical parameters from oscillating time series
Ausgabe der Zeitschrift
17

Datenquelle: Crossref

Abstract
Deep neural networks are widely used in pattern-recognition tasks for which a human-comprehensible, quantitative description of the data-generating process, cannot be obtained. While doing so, neural networks often produce an abstract (entangled and non-interpretable) representation of the data-generating process. This may be one of the reasons why neural networks are not yet used extensively in physics-experiment signal processing: physicists generally require their analyses to yield quantitative information about the system they study. In this article we use a deep neural network to disentangle components of oscillating time series. To this aim, we design and train the neural network on synthetic oscillating time series to perform two tasks: a regression of the signal latent parameters and signal denoising by an Autoencoder-like architecture. We show that the regression and denoising performance is similar to those of least-square curve fittings with true latent-parameters initial guesses, in spite of the neural network needing no initial guesses at all. We then explore various applications in which we believe our architecture could prove useful for time-series processing, when prior knowledge is incomplete. As an example, we employ the neural network as a preprocessing tool to inform the least-square fits when initial guesses are unknown. Moreover, we show that the regression can be performed on some latent parameters, while ignoring the existence of others. Because the Autoencoder needs no prior information about the physical model, the remaining unknown latent parameters can still be captured, thus making use of partial prior knowledge, while leaving space for data exploration and discoveries.
Addresses
  • Johannes Gutenberg-Universität, Mainz, Germany.
Autoren
  • Antoine Garcon
  • Julian Vexler
  • Dmitry Budker
  • Stefan Kramer
DOI
10.1371/journal.pone.0268439
eISSN
1932-6203
Externe Identifier
  • PubMed Identifier: 35560322
  • PubMed Central ID: PMC9106171
Funding acknowledgements
  • carl-zeiss-stiftung:
  • European Research Council: 695405
  • german research foundation: 39083149
  • dfg, reinhart koselleck project:
Open access
true
ISSN
1932-6203
Ausgabe der Veröffentlichung
5
Zeitschrift
PloS one
Schlüsselwörter
  • Humans
  • Physics
  • Knowledge
  • Time Factors
  • Signal Processing, Computer-Assisted
  • Neural Networks, Computer
Sprache
eng
Medium
Electronic-eCollection
Online publication date
2022
Open access status
Open Access
Paginierung
e0268439
Datum der Veröffentlichung
2022
Status
Published
Publisher licence
CC BY
Datum der Datenerfassung
2022
Titel
Deep neural networks to recover unknown physical parameters from oscillating time series.
Sub types
  • Research Support, Non-U.S. Gov't
  • research-article
  • Journal Article
Ausgabe der Zeitschrift
17

Files

https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0268439&type=printable https://europepmc.org/articles/PMC9106171?pdf=render

Datenquelle: Europe PubMed Central

Abstract
Deep neural networks are widely used in pattern-recognition tasks for which a human-comprehensible, quantitative description of the data-generating process, cannot be obtained. While doing so, neural networks often produce an abstract (entangled and non-interpretable) representation of the data-generating process. This may be one of the reasons why neural networks are not yet used extensively in physics-experiment signal processing: physicists generally require their analyses to yield quantitative information about the system they study. In this article we use a deep neural network to disentangle components of oscillating time series. To this aim, we design and train the neural network on synthetic oscillating time series to perform two tasks: a regression of the signal latent parameters and signal denoising by an Autoencoder-like architecture. We show that the regression and denoising performance is similar to those of least-square curve fittings with true latent-parameters initial guesses, in spite of the neural network needing no initial guesses at all. We then explore various applications in which we believe our architecture could prove useful for time-series processing, when prior knowledge is incomplete. As an example, we employ the neural network as a preprocessing tool to inform the least-square fits when initial guesses are unknown. Moreover, we show that the regression can be performed on some latent parameters, while ignoring the existence of others. Because the Autoencoder needs no prior information about the physical model, the remaining unknown latent parameters can still be captured, thus making use of partial prior knowledge, while leaving space for data exploration and discoveries.
Date of acceptance
2022
Autoren
  • Antoine Garcon
  • Julian Vexler
  • Dmitry Budker
  • Stefan Kramer
Autoren-URL
https://www.ncbi.nlm.nih.gov/pubmed/35560322
DOI
10.1371/journal.pone.0268439
eISSN
1932-6203
Externe Identifier
  • PubMed Central ID: PMC9106171
Ausgabe der Veröffentlichung
5
Zeitschrift
PLoS One
Schlüsselwörter
  • Humans
  • Knowledge
  • Neural Networks, Computer
  • Physics
  • Signal Processing, Computer-Assisted
  • Time Factors
Sprache
eng
Country
United States
Paginierung
e0268439
PII
PONE-D-21-24201
Datum der Veröffentlichung
2022
Status
Published online
Datum, an dem der Datensatz öffentlich gemacht wurde
2022
Titel
Deep neural networks to recover unknown physical parameters from oscillating time series.
Sub types
  • Journal Article
  • Research Support, Non-U.S. Gov't
Ausgabe der Zeitschrift
17

Datenquelle: PubMed

Beziehungen:
Hat Preprint
Eigentum von

Werkzeuge