TASE-Stabilized Time-Evolving Natural Gradient Methods for Diffusion-Dominated PDEs

DOI10.48550/arXiv.2609.23054MaRDI QIDQ7353453arXivFDO

Authors Zihao Shi, Dongling Wang

Publication date 19 September 2026

Published in arXiv

Tldr We propose a new method TASE-TENG that combines Time-Evolving Natural Gradient (TENG) methods for neural approximations of time-dependent PDEs with a highly stable explicit operator, which is constructed from a learnable surrogate operator and improves stability in stiff diffusion-dominated problems while retaining accuracy.

Full work available at URL https://doi.org/10.48550/arXiv.2609.23054

arXiv classification

math.NA

Cites work

• A note on the stability of time-accurate and highly-stable explicit operators for stiff differential equations
• A Stable and Scalable Method for Solving Initial Value PDEs with Neural Networks
• Characterizing possible failure modes in physics-informed neural networks
• Enforcing exact physics in scientific machine learning: a data-driven exterior calculus on graphs
• From quantum to classical molecular dynamics: Reduced models and numerical analysis.
• Metastability as a coexistence mechanism in a model for dryland vegetation patterns
• Parsimonious Physics-Informed Random Projection Neural Networks for Initial-Value Problems of ODEs and index-1 DAEs
• Preserving conservation laws in the time-evolving natural Galerkin method via relaxation and projection
• Randomized Sparse Neural Galerkin Schemes for Solving Evolution Equations with Deep Networks
• Randomized time stepping of nonlinearly parametrized solutions of evolution problems
• Solving multiscale dynamical systems by deep learning
• Stabilize physics-informed neural networks for stiff differential equations: re-spacing layer
• Time-Accurate and highly-Stable Explicit operators for stiff differential equations
• Understanding and Mitigating Gradient Flow Pathologies in Physics-Informed Neural Networks

This page was built for publication: TASE-Stabilized Time-Evolving Natural Gradient Methods for Diffusion-Dominated PDEs

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7353453)