Score-Based Stabilization for Time-Dependent Problems

DOI10.48550/arXiv.2607.25119MaRDI QIDQ7284455arXivFDO

Authors Eldad Haber, Uri M. Ascher, Eshed Gal, Uri Ascher

Publication date 27 July 2026

Published in arXiv

Tldr We propose a score-based stabilization framework for numerical simulation of partial differential equations by augmenting standard time-stepping schemes with a learned score model that defines a stabilization operator acting as a contraction toward the manifold of admissible states. This operator yields a correction mechanism with basin-conditional stability, demonstrated to improve robustness and suppress nonphysical instabilities in Advection, KdV, NLS, and Burgers' equations.

Full work available at URL https://doi.org/10.48550/arXiv.2607.25119

arXiv classification

cs.LGmath.NA

Cites work

• A Method for the Numerical Calculation of Hydrodynamic Shocks
• Adam: A Method for Stochastic Optimization
• Efficient implementation of weighted ENO schemes
• Estimation of non-normalized statistical models by score matching
• Estimation of ordinary differential equation models with discretization error quantification
• Flux-corrected transport. I: SHASTA, a fluid transport algorithm that works
• Geometric Numerical Integration
• Geometric numerical integration and Schrödinger equations
• High resolution schemes for hyperbolic conservation laws
• Interaction of `solitons' in a collisionless plasma and the recurrence of initial states
• Learning data-driven discretizations for partial differential equations
• On splitting methods for Schrödinger-Poisson and cubic nonlinear Schrödinger equations
• On symplectic and multisymplectic schemes for the KdV equation
• Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations
• Quantifying uncertainty in the numerical integration of evolution equations based on Bayesian isotonic regression
• Spectral methods for hyperbolic problems
• Split-Step Methods for the Solution of the Nonlinear Schrödinger Equation
• Tweedie’s Formula and Selection Bias
• U-Net: Convolutional Networks for Biomedical Image Segmentation

This page was built for publication: Score-Based Stabilization for Time-Dependent Problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7284455)