Non-Asymptotic Analysis of Projected Gradient Descent for Physics-Informed Neural Networks

Byadmin
  • Authors: Jonas Nießen, Johannes Müller
  • First public date: 2025-05-12
  • arXiv: 2505.07311
  • Preprint year: 2025
  • Status: Published
  • Publication type: Book chapter
  • Publication year: 2026
  • Book: Scientific Machine Learning: Emerging Topics, SEMA SIMAI Springer Series (2026)
  • DOI: 10.1007/978-3-032-11527-0_4

Abstract

In this work, we provide a non-asymptotic convergence analysis of projected gradient descent for physics-informed neural networks for the Poisson equation. Under suitable assumptions, we show that the optimization error can be bounded by $\mathcal{O}(1/\sqrt{T} + 1/\sqrt{m} + ε_{\text{approx}})$, where $T$ is the number of algorithm time steps, $m$ is the width of the neural network and $ε_{\text{approx}}$ is an approximation error. The proof of our optimization result relies on bounding the linearization error and using this result together with a Lyapunov drift analysis. Additionally, we quantify the generalization error by bounding the Rademacher complexities of the neural network and its Laplacian. Combining both the optimization and generalization results, we obtain an overall error estimate based on an existing error estimate from regularity theory.

Associated SAiS members

Research areas

BibTeX

@incollection{arxiv250507311,
  title = {Non-Asymptotic Analysis of Projected Gradient Descent for Physics-Informed Neural Networks},
  author = {Jonas Nießen and Johannes Müller},
  year = {2026},
  booktitle = {Scientific Machine Learning: Emerging Topics, SEMA SIMAI Springer Series (2026)},
  doi = {10.1007/978-3-032-11527-0_4},
  eprint = {2505.07311},
  archivePrefix = {arXiv},
  primaryClass = {math.OC}
}

Similar Posts