Salvato in:
Dettagli Bibliografici
Autori principali: Rojas, Sergio, Maczuga, Paweł, Muñoz-Matute, Judit, Pardo, David, Paszynski, Maciej
Natura: Preprint
Pubblicazione: 2023
Soggetti:
Accesso online:https://arxiv.org/abs/2308.16910
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866916147118473216
author Rojas, Sergio
Maczuga, Paweł
Muñoz-Matute, Judit
Pardo, David
Paszynski, Maciej
author_facet Rojas, Sergio
Maczuga, Paweł
Muñoz-Matute, Judit
Pardo, David
Paszynski, Maciej
contents We introduce a Robust version of the Variational Physics-Informed Neural Networks method (RVPINNs). As in VPINNs, we define the quadratic loss functional in terms of a Petrov-Galerkin-type variational formulation of the PDE problem: the trial space is a (Deep) Neural Network (DNN) manifold, while the test space is a finite-dimensional vector space. Whereas the VPINN's loss depends upon the selected basis functions of a given test space, herein, we minimize a loss based on the discrete dual norm of the residual. The main advantage of such a loss definition is that it provides a reliable and efficient estimator of the true error in the energy norm under the assumption of the existence of a local Fortin operator. We test the performance and robustness of our algorithm in several advection-diffusion problems. These numerical results perfectly align with our theoretical findings, showing that our estimates are sharp.
format Preprint
id arxiv_https___arxiv_org_abs_2308_16910
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Robust Variational Physics-Informed Neural Networks
Rojas, Sergio
Maczuga, Paweł
Muñoz-Matute, Judit
Pardo, David
Paszynski, Maciej
Numerical Analysis
65N12, 65N15, 65N22, 65N30, 65N50
We introduce a Robust version of the Variational Physics-Informed Neural Networks method (RVPINNs). As in VPINNs, we define the quadratic loss functional in terms of a Petrov-Galerkin-type variational formulation of the PDE problem: the trial space is a (Deep) Neural Network (DNN) manifold, while the test space is a finite-dimensional vector space. Whereas the VPINN's loss depends upon the selected basis functions of a given test space, herein, we minimize a loss based on the discrete dual norm of the residual. The main advantage of such a loss definition is that it provides a reliable and efficient estimator of the true error in the energy norm under the assumption of the existence of a local Fortin operator. We test the performance and robustness of our algorithm in several advection-diffusion problems. These numerical results perfectly align with our theoretical findings, showing that our estimates are sharp.
title Robust Variational Physics-Informed Neural Networks
topic Numerical Analysis
65N12, 65N15, 65N22, 65N30, 65N50
url https://arxiv.org/abs/2308.16910