EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Can physics-informed neural networks beat the finite element method?

Tamara G Grossmann, Urszula Julia Komorowska, Jonas Latz iD, Carola-Bibiane Schönlieb iD

IMA Journal of Applied Mathematics · 21.06.2024 · 89 (1) · 143-174

DOI: 10.1093/imamat/hxae011 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IMA Journal of Applied Mathematics, online erschienen am 23.05.2024. Verfasst von Tamara G Grossmann und 3 weiteren Autor:innen. Beteiligte Einrichtung: Department of Applied Mathematics and Theoretical Physics, University of Cambridge, Wilberforce Road, Cambridge CB3 0WA, UK. Die Arbeit wurde bisher 220-mal zitiert (Crossref) und nennt 81 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Cantab Capital Institute for the Mathematics of Information, European Union Horizon 2020, EPSRC National Productivity and Investment Fund.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Partial differential equations (PDEs) play a fundamental role in the mathematical modelling of many processes and systems in physical, biological and other sciences. To simulate such processes and systems, the solutions of PDEs often need to be approximated numerically. The finite element method, for instance, is a usual standard methodology to do so. The recent success of deep neural networks at various approximation tasks has motivated their use in the numerical solution of PDEs. These so-called physics-informed neural networks and their variants have shown to be able to successfully approximate a large range of PDEs. So far, physics-informed neural networks and the finite element method have mainly been studied in isolation of each other. In this work, we compare the methodologies in a systematic computational study. Indeed, we employ both methods to numerically solve various linear and nonlinear PDEs: Poisson in 1D, 2D and 3D, Allen–Cahn in 1D, semilinear Schrödinger in 1D and 2D. We then compare computational costs and approximation accuracies. In terms of solution time and accuracy, physics-informed neural networks have not been able to outperform the finite element method in our study. In some experiments, they were faster at evaluating the solved PDE.

Schlagwörter

Stichwörter der Autor:innen

deep learningfinite element methodpartial differential equationsphysics-informed neural networks

Autor:innen und Institutionen

  • Department of Applied Mathematics and Theoretical Physics, University of Cambridge, Wilberforce Road, Cambridge CB3 0WA, UK.
  • Department of Computer Science and Technology, University of Cambridge, 15 JJ Thomson Avenue, Cambridge CB3 0FD, UK.
  • Department of Mathematics, University of Manchester, Alan Turing Building, Oxford Road, Manchester M13 9PL, UK.

ORCID-Profile: Jonas Latz, Carola-Bibiane Schönlieb

Förderung

  • Cantab Capital Institute for the Mathematics of Information
  • European Union Horizon 2020
  • EPSRC National Productivity and Investment Fund
  • EPSRC
  • Philip Leverhulme Prize
  • Royal Society Wolfson Fellowship
  • Wellcome Trust
  • Alan Turing Institute

Zitieren

APA

Grossmann, T. G., Komorowska, U. J., Latz, J., & Schönlieb, C. B. (2024). Can physics-informed neural networks beat the finite element method?. IMA Journal of Applied Mathematics, 89(1), 143-174. https://doi.org/10.1093/imamat/hxae011

Vancouver

Grossmann TG, Komorowska UJ, Latz J, Schönlieb CB. Can physics-informed neural networks beat the finite element method?. IMA Journal of Applied Mathematics. 2024;89(1):143-174. doi:10.1093/imamat/hxae011

BibTeX
@article{Grossmann2024,
  author = {Grossmann, Tamara G and Komorowska, Urszula Julia and Latz, Jonas and Schönlieb, Carola-Bibiane},
  title = {Can physics-informed neural networks beat the finite element method?},
  journal = {IMA Journal of Applied Mathematics},
  year = {2024},
  volume = {89},
  number = {1},
  pages = {143--174},
  doi = {10.1093/imamat/hxae011}
}
Kostenlos starten →