analiza aiqbench
The paper argues that test loss on solver-generated data can't tell a learned physics model from one that has absorbed the solver's discretization error, so evaluating ML surrogates for PDEs needs other checks. Its proposed diagnostics test whether a trained model respects a known invariant, symmetry, or Hamiltonian structure, without needing training data or the governing equation. The reported rank correlations with true error (up to 0.94) are promising, but they come from five systems, so readers should watch whether the diagnostics generalize to other PDEs and architectures.
Structure-preserving fidelity diagnostics for partial differential equations
arXiv:2610.17483v1 Announce Type: new 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.
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.
Machine Learning (cs.LG); Computational Physics (physics.comp-ph)
arXiv:2610.17483 [cs.LG]
(or arXiv:2610.17483v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.17483 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 272 cuvinte
Subiecte conexe
Predicting Out-of-Distribution Generalization of Neural Operators via Observable Spectral Error Decomposition
arXiv · 22 sept. 2026
Researchers Link Sensor Count to Data Accuracy in Simulations
Quantum Zeitgeist · 2 oct. 2026
Performance Profiling on AMD GPUs – Part 5: Profiling-Driven Kernel Optimization with an AI Code-Assist Tool
AMD · 16 iul. 2026
Learning PDE solution operators with variable initial conditions via Latent Dynamics Networks
arXiv · 7 oct. 2026