Learning the Entropic Relaxation: Stable Coarse-Grid Emulation with Neural Discrete Equilibrium
Abstract
The ability to emulate complicated and computationally-expensive physical systems in an efficient and stable manner is a popular motivation for scientific machine learning (SciML). Stable numerical solvers tend to be computationally expensive, as they attempt to resolve small (spatial and temporal) scales that receive and dissipate spectral energy. Learned coarse-grid forecasters remove the dissipative channel, allowing unresolved frequencies to re-emerge as grid-scale noise (aliasing). In this paper, we use the Neural Discrete Equilibrium (NeurDE) method to explore stable emulation at extremely coarse grids, demonstrating physically-principled control over entropy dynamics that reinforces the physical system's dissipative mechanism. Entropically-stabilized NeurDE produces stable long-time rollouts for thousands of forecasting steps on large Reynolds number turbulent flows (Re=100,000) with very coarse grids (down to ℝ16×16). At these very coarse grid scales, entropically-stabilized NeurDE can be more stable than even our stabilized data-generation numerical method.
Related
- Tight Quasiprobabilistic Entropic Uncertainty Relations for Discrete Quantum Systems
- An entropic barrier to equilibration in the discrete nonlinear Schrödinger equation: Rigorous bounds for the microcanonical Brownian motion
- Out-of-equilibrium relaxation dynamics of the superconducting order parameter in CsV3Sb5
- On the relaxation dynamics of non-equilibrium quantum systems
- Equivalence classes of finite-time transitions in optimal control and non-equilibrium relaxation
BibTeX
@misc{benitez2026learning,
title = {{Learning the Entropic Relaxation: Stable Coarse-Grid Emulation with Neural Discrete Equilibrium}},
author = {Jose Antonio Lara Benitez and Kareem Hegazy and Anastasis Kratsios and Ivan Dokmanić and Maarten V. de Hoop and Michael W. Mahoney},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.02458},
eprint = {2611.02458},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.02458},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Learning the Entropic Relaxation: Stable Coarse-Grid Emulation with Neural Discrete Equilibrium
AU - Jose Antonio Lara Benitez
AU - Kareem Hegazy
AU - Anastasis Kratsios
AU - Ivan Dokmanić
AU - Maarten V. de Hoop
AU - Michael W. Mahoney
PY - 2026
DA - 2026/11/02
PB - arXiv
DO - 10.48550/arxiv.2611.02458
UR - https://arxiv.org/abs/2611.02458
AB - The ability to emulate complicated and computationally-expensive physical systems in an efficient and stable manner is a popular motivation for scientific machine learning (SciML). Stable numerical solvers tend to be computationally expensive, as they attempt to resolve small (spatial and temporal) scales that receive and dissipate spectral energy. Learned coarse-grid forecasters remove the dissipative channel, allowing unresolved frequencies to re-emerge as grid-scale noise (aliasing). In this paper, we use the Neural Discrete Equilibrium (NeurDE) method to explore stable emulation at extremely coarse grids, demonstrating physically-principled control over entropy dynamics that reinforces the physical system's dissipative mechanism. Entropically-stabilized NeurDE produces stable long-time rollouts for thousands of forecasting steps on large Reynolds number turbulent flows (Re=100,000) with very coarse grids (down to ℝ^16×16). At these very coarse grid scales, entropically-stabilized NeurDE can be more stable than even our stabilized data-generation numerical method.
ER -