EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Data-driven, ML-assisted approaches to problem well-posedness

Tom Bertalan, George A Kevrekidis iD, Eleni D Koronaki, Siddhartha Mishra, Elizaveta Rebrova iD, Ioannis G Kevrekidis iD

PNAS Nexus · 30.05.2026 · 5 (6)

DOI: 10.1093/pnasnexus/pgag175 ↗

Kurz gefasst

Fachartikel in der Zeitschrift PNAS Nexus, online erschienen am 01.06.2026. Verfasst von Tom Bertalan und 5 weiteren Autor:innen. Beteiligte Einrichtung: Transformative Digital Capabilities, Amgen, Cambridge, MA 02142, USA. Die Arbeit nennt 26 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0). Gefördert von: Department of Energy and by the US National Science Foundation, NSF.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Classically, to solve differential equation problems, it is necessary to specify sufficient initial condition and/or boundary condition (BC) so as to allow the existence of a unique solution. Well-posedness of differential equation problems thus involves studying the existence and uniqueness of solutions, and their dependence on such prespecified conditions. However, in part due to mathematical necessity, these conditions are usually specified “to arbitrary precision” only on (appropriate portions of) the boundary of the space-time domain. This does not mirror how data acquisition is performed in realistic situations, where one may observe entire “patches” of solution data at arbitrary space-time locations; alternatively one might have access to more than one solution stemming from the same differential operator. In our short work, we demonstrate how standard tools from machine and manifold learning can be used to infer, in a data driven manner, certain well-posedness features of differential equation problems, for initial condition/BC combinations under which rigorous existence/uniqueness theorems are not known. Our study naturally combines a data assimilation perspective with an operator-learning one.

Schlagwörter

Stichwörter der Autor:innen

differential equationsphysics-informed neural networksrandomized linear algebrasolution manifoldswell-posedness

Autor:innen und Institutionen

  • Transformative Digital Capabilities, Amgen, Cambridge, MA 02142, USA.
  • Department of Applied Mathematics and Statistics, Johns Hopkins University, Baltimore, MD 21218, USA.
  • Los Alamos National Laboratory, Los Alamos, NM 87545, USA.
  • Faculty of Science, Technology and Medicine, University of Luxembourg, Esch-sur-Alzette L-4365, Luxembourg.
  • Seminar for Applied Mathematics, Department of Mathematics, and ETH AI Center, ETH Zurich, 8092 Zurich, Switzerland.
  • Department of Operations Research and Financial Engineering, Princeton University, Princeton, NJ 08544, USA.

ORCID-Profile: George A Kevrekidis, Elizaveta Rebrova, Ioannis G Kevrekidis

Förderung

  • Department of Energy and by the US National Science Foundation
  • NSF

Zitieren

APA

Bertalan, T., Kevrekidis, G. A., Koronaki, E. D., Mishra, S., Rebrova, E., & Kevrekidis, I. G. (2026). Data-driven, ML-assisted approaches to problem well-posedness. PNAS Nexus, 5(6). https://doi.org/10.1093/pnasnexus/pgag175

Vancouver

Bertalan T, Kevrekidis GA, Koronaki ED, Mishra S, Rebrova E, Kevrekidis IG. Data-driven, ML-assisted approaches to problem well-posedness. PNAS Nexus. 2026;5(6). doi:10.1093/pnasnexus/pgag175

BibTeX
@article{Bertalan2026,
  author = {Bertalan, Tom and Kevrekidis, George A and Koronaki, Eleni D and Mishra, Siddhartha and Rebrova, Elizaveta and Kevrekidis, Ioannis G},
  title = {Data-driven, ML-assisted approaches to problem well-posedness},
  journal = {PNAS Nexus},
  year = {2026},
  volume = {5},
  number = {6},
  doi = {10.1093/pnasnexus/pgag175}
}
Kostenlos starten →