A simple remedy for failure modes in physics informed neural networks

DOI10.1016/J.NEUNET.2024.106963MaRDI QIDQ6670022zbMATH OpenFDO

Authors Ghazal Farhani, Nima Hosseini Dashtbayaz, Alexander Kazachek, Boyu Wang

Publication date 22 January 2025

Published in Neural Networks (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1016/J.NEUNET.2024.106963

zbMATH Keywords

failure modes in PINNsoptimizationphysics-informed neural networks

Mathematics Subject Classification ID

Partial differential equations (35-XX) Numerical analysis (65-XX)

Recommendations

• Investigating and Mitigating Failure Modes in Physics-Informed Neural Networks (PINNs)
• When and why PINNs fail to train: a neural tangent kernel perspective
• An analysis and solution of ill-conditioning in physics-informed neural networks
• Less Emphasis on Hard Regions: Curriculum Learning of PINNs for Singularly Perturbed Convection-Diffusion-Reaction Problems
• Navigating PINNs via maximum residual-based continuous distribution

Cited in

(5)

• When and why PINNs fail to train: a neural tangent kernel perspective
• Understanding and Mitigating Gradient Flow Pathologies in Physics-Informed Neural Networks
• Less Emphasis on Hard Regions: Curriculum Learning of PINNs for Singularly Perturbed Convection-Diffusion-Reaction Problems
• Loss jump during loss switch in solving PDEs with neural networks
• An analysis and solution of ill-conditioning in physics-informed neural networks

This page was built for publication: A simple remedy for failure modes in physics informed neural networks

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6670022)