Numerical solution of Poisson partial differential equation in high dimension using two-layer neural networks - Archive ouverte HAL Accéder directement au contenu
Pré-Publication, Document De Travail Année : 2023

Numerical solution of Poisson partial differential equation in high dimension using two-layer neural networks

Résumé

The aim of this article is to analyze numerical schemes using two-layer neural networks with infinite width for the resolution of the high-dimensional Poisson-Neumann partial differential equations (PDEs) with Neumann boundary conditions. Using Barron's representation of the solution [1] with a measure of probability, the energy is minimized thanks to a gradient curve dynamic on the 2 Wasserstein space of parameters defining the neural network. Inspired by the work from Bach and Chizat [2, 3], we prove that if the gradient curve converges, then the represented function is the solution of the elliptic equation considered. In contrast to the works [2, 3], the activation function we use here is not assumed to be homogeneous to obtain global convergence of the flow. Numerical experiments are given to show the potential of the method.
Fichier principal
Vignette du fichier
main.pdf (779.23 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04089961 , version 1 (05-05-2023)
hal-04089961 , version 2 (13-07-2023)

Licence

Paternité - Partage selon les Conditions Initiales

Identifiants

  • HAL Id : hal-04089961 , version 2

Citer

Mathias Dus, Virginie Ehrlacher. Numerical solution of Poisson partial differential equation in high dimension using two-layer neural networks. 2023. ⟨hal-04089961v2⟩
90 Consultations
127 Téléchargements

Partager

Gmail Facebook X LinkedIn More