Enhancing Training of Physics-Informed Neural Networks Using Domain Decomposition–Based Preconditioning Strategies
SIAM Journal on Scientific Computing · 31.10.2024 · 46 (5) · S46-S67
DOI: 10.1137/23m1583375 ↗Kurz gefasst
Fachartikel in der Zeitschrift SIAM Journal on Scientific Computing, online erschienen am 11.04.2024. Verfasst von Alena Kopaničáková und 3 weiteren Autor:innen. Die Arbeit wurde bisher 32-mal zitiert (Crossref) und nennt 79 Literaturangaben. Gefördert von: Swiss National Science Foundation, Platform for Advanced Scientific Computing, Vannevar Bush Faculty Fellowship.
Abstract (Originalsprache)
We propose to enhance the training of physics-informed neural networks. To this aim, we introduce nonlinear additive and multiplicative preconditioning strategies for the widely used L-BFGS optimizer. The nonlinear preconditioners are constructed by utilizing the Schwarz domain decomposition framework, where the parameters of the network are decomposed in a layerwise manner. Through a series of numerical experiments, we demonstrate that both additive and multiplicative preconditioners significantly improve the convergence of the standard L-BFGS optimizer while providing more accurate solutions of the underlying PDEs. Moreover, the additive preconditioner is inherently parallel, thus giving rise to a novel approach to model parallelism.
Förderung
- Swiss National Science Foundation
- Platform for Advanced Scientific Computing
- Vannevar Bush Faculty Fellowship
Zitieren
APA
Kopaničáková, A., Kothari, H., Karniadakis, G. E., & Krause, R. (2024). Enhancing Training of Physics-Informed Neural Networks Using Domain Decomposition–Based Preconditioning Strategies. SIAM Journal on Scientific Computing, 46(5), S46-S67. https://doi.org/10.1137/23m1583375
Vancouver
Kopaničáková A, Kothari H, Karniadakis GE, Krause R. Enhancing Training of Physics-Informed Neural Networks Using Domain Decomposition–Based Preconditioning Strategies. SIAM Journal on Scientific Computing. 2024;46(5):S46-S67. doi:10.1137/23m1583375
BibTeX
@article{Kopanikov2024,
author = {Kopaničáková, Alena and Kothari, Hardik and Karniadakis, George E. and Krause, Rolf},
title = {Enhancing Training of Physics-Informed Neural Networks Using Domain Decomposition–Based Preconditioning Strategies},
journal = {SIAM Journal on Scientific Computing},
year = {2024},
volume = {46},
number = {5},
pages = {S46--S67},
doi = {10.1137/23m1583375}
}
EUVIMED