Structure-preserving fidelity diagnostics for partial differential equations
Abstract
Machine learning surrogates of partial differential equations (PDEs) are commonly trained on data generated by numerical solvers. Therefore, the training data obey not the PDE but the modified equation of the numerical scheme, and the test loss, computed on data from the same solver, cannot distinguish a model that learned the physics from one that learned the discretization error. Here, we show that the difference is nevertheless visible in the trained model. We propose a suite of structure-preserving fidelity diagnostics, simple scores that measure whether a learned model respects one known property of the physics, such as a conserved quantity, a symmetry or a Hamiltonian structure, without access to the training data or the governing equation. Over five systems, the diagnostics track the error with respect to the true dynamics with a rank correlation of up to $0.94$, while the test loss is uncorrelated or anti-correlated with it. For the Burgers, Korteweg-de Vries and Navier-Stokes equations, the diagnostics separate numerical schemes that the test loss ranks as equivalent, including schemes of the same order. order.
Related
- Structure-Preserving Full Approximation Scheme for Real-Space Hartree-Fock Equations
- Fast and High-Fidelity Motional-State-Preserving Transport in Optical Tweezers
- On Bäcklund transformations preserving the Darboux integrability of hyperbolic equations
- Bayesian data-driven discovery of partial differential equations with variable coefficients
- Fast quantum algorithm for differential equations
BibTeX
@misc{flouris2026structurepreserving,
title = {{Structure-preserving fidelity diagnostics for partial differential equations}},
author = {Kyriakos Flouris},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.17483},
eprint = {2610.17483},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.17483},
note = {CC0},
}
RIS
TY - UNPB
TI - Structure-preserving fidelity diagnostics for partial differential equations
AU - Kyriakos Flouris
PY - 2026
DA - 2026/10/12
PB - arXiv
DO - 10.48550/arxiv.2610.17483
UR - https://arxiv.org/abs/2610.17483
AB - Machine learning surrogates of partial differential equations (PDEs) are commonly trained on data generated by numerical solvers. Therefore, the training data obey not the PDE but the modified equation of the numerical scheme, and the test loss, computed on data from the same solver, cannot distinguish a model that learned the physics from one that learned the discretization error. Here, we show that the difference is nevertheless visible in the trained model. We propose a suite of structure-preserving fidelity diagnostics, simple scores that measure whether a learned model respects one known property of the physics, such as a conserved quantity, a symmetry or a Hamiltonian structure, without access to the training data or the governing equation. Over five systems, the diagnostics track the error with respect to the true dynamics with a rank correlation of up to $0.94$, while the test loss is uncorrelated or anti-correlated with it. For the Burgers, Korteweg-de Vries and Navier-Stokes equations, the diagnostics separate numerical schemes that the test loss ranks as equivalent, including schemes of the same order. order.
ER -