EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Convergence of physics-informed neural networks for time-fractional diffusion equations via stability and generalization bounds

Phuoc Toan Trinh, Huu Dinh Huynh

Electronic Journal of Applied Mathematics · 26.12.2025 · 3 (4) · 35-55

DOI: 10.61383/ejam.202534116 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Electronic Journal of Applied Mathematics, online erschienen am 26.12.2025. Verfasst von Phuoc Toan Trinh und 1 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 18 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Physics-informed neural networks (PINNs) have emerged as a flexible framework for solving partial differential equations (PDEs), yet rigorous convergence results remain limited for models with nonlocal time dependence. This paper studies PINNs for the time-fractional diffusion equation with the Caputo derivative of order \(\alpha\in(0,1)\). We propose a residual-based PINN formulation and establish a stability-driven error bound: the solution error is controlled by the PDE residual together with the boundary and initial condition residuals. By combining this PDE stability with sampling-based generalization bounds for the empirical PINN loss, we prove convergence of empirical minimizers to the true solution as the number of collocation points increases and the approximation and optimization errors vanish. Numerical experiments with manufactured solutions support the theory and demonstrate accuracy.

Zitieren

APA

Trinh, P. T., & Huynh, H. D. (2025). Convergence of physics-informed neural networks for time-fractional diffusion equations via stability and generalization bounds. Electronic Journal of Applied Mathematics, 3(4), 35-55. https://doi.org/10.61383/ejam.202534116

Vancouver

Trinh PT, Huynh HD. Convergence of physics-informed neural networks for time-fractional diffusion equations via stability and generalization bounds. Electronic Journal of Applied Mathematics. 2025;3(4):35-55. doi:10.61383/ejam.202534116

BibTeX
@article{Trinh2025,
  author = {Trinh, Phuoc Toan and Huynh, Huu Dinh},
  title = {Convergence of physics-informed neural networks for time-fractional diffusion equations via stability and generalization bounds},
  journal = {Electronic Journal of Applied Mathematics},
  year = {2025},
  volume = {3},
  number = {4},
  pages = {35--55},
  doi = {10.61383/ejam.202534116}
}
Kostenlos starten →