Discontinuous hybrid neural networks for the one-dimensional partial differential equations

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wang, Xiaoyu, Yuan, Long, Yu, Yao
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913838679457792
author Wang, Xiaoyu
Yuan, Long
Yu, Yao
author_facet Wang, Xiaoyu
Yuan, Long
Yu, Yao
contents A feedforward neural network, including hidden layers, motivated by nonlinear functions (such as Tanh, ReLU, and Sigmoid functions), exhibits uniform approximation properties in Sobolev space, and discontinuous neural networks can reduce computational complexity. In this work, we present a discontinuous hybrid neural network method for solving the partial differential equations, construct a new hybrid loss functional that incorporates the variational of the approximation equation, interface jump stencil and boundary constraints. The RMSprop algorithm and discontinuous Galerkin method are employed to update the nonlinear parameters and linear parameters in neural networks, respectively. This approach guarantees the convergence of the loss functional and provides an approximate solution with high accuracy.
format Preprint
id arxiv_https___arxiv_org_abs_2505_09911
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Discontinuous hybrid neural networks for the one-dimensional partial differential equations
Wang, Xiaoyu
Yuan, Long
Yu, Yao
Numerical Analysis
65N30, 65N55, 68T07
A feedforward neural network, including hidden layers, motivated by nonlinear functions (such as Tanh, ReLU, and Sigmoid functions), exhibits uniform approximation properties in Sobolev space, and discontinuous neural networks can reduce computational complexity. In this work, we present a discontinuous hybrid neural network method for solving the partial differential equations, construct a new hybrid loss functional that incorporates the variational of the approximation equation, interface jump stencil and boundary constraints. The RMSprop algorithm and discontinuous Galerkin method are employed to update the nonlinear parameters and linear parameters in neural networks, respectively. This approach guarantees the convergence of the loss functional and provides an approximate solution with high accuracy.
title Discontinuous hybrid neural networks for the one-dimensional partial differential equations
topic Numerical Analysis
65N30, 65N55, 68T07
url https://arxiv.org/abs/2505.09911