Machine-learned spectral PDE solvers for fluids and plasmas
Abstract
Sparse regression methods for discovering governing partial differential equations (PDEs) from data are sensitive to observation noise and are not constrained to reproduce the data under forward integration. We address both limitations by performing weak-form discovery in the spatial discretization in which the PDE is solved, learned by differentiating through multistep rollouts anchored at the observed states. The new method, Weak-form Anchored Rollout for PDEs (WARP), produces a machine-learned PDE solver. In this work, we focus on systems amenable to a spectral Galerkin solver. Across eight benchmark fluid and plasma systems, including high-dimensional turbulent simulations, WARP produces numerically stable rollouts from new random initial conditions and substantially improves coefficient recovery, rollout accuracy, and statistical measures compared with a single-step weak-form SINDy baseline.
Related
- GENIE: learned quantum circuits as embedding impurity solvers, and why compactness, not correlation strength, decides when they work
- Accessing the fundamental link between atomic populations and frequency distribution of spectral lines of ions in dense plasmas
- Repairability of Inexact Solvers in Recursive State Estimation with Machine Learning
- Validation of geometric proxies using negative and positive triangularity plasmas with varying elongation on TCV and multi-machine gyrokinetic simulations
- Multivariate hyperdensity functional theory for inhomogeneous equilibrium fluids: From first principles to simulation-based machine learning
BibTeX
@misc{chen2026machinelearned,
title = {{Machine-learned spectral PDE solvers for fluids and plasmas}},
author = {Mellow Baixuan Chen and Alan A. Kaptanoglu},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.27317},
eprint = {2610.27317},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.27317},
}
RIS
TY - UNPB
TI - Machine-learned spectral PDE solvers for fluids and plasmas
AU - Mellow Baixuan Chen
AU - Alan A. Kaptanoglu
PY - 2026
DA - 2026/10/19
PB - arXiv
DO - 10.48550/arxiv.2610.27317
UR - https://arxiv.org/abs/2610.27317
AB - Sparse regression methods for discovering governing partial differential equations (PDEs) from data are sensitive to observation noise and are not constrained to reproduce the data under forward integration. We address both limitations by performing weak-form discovery in the spatial discretization in which the PDE is solved, learned by differentiating through multistep rollouts anchored at the observed states. The new method, Weak-form Anchored Rollout for PDEs (WARP), produces a machine-learned PDE solver. In this work, we focus on systems amenable to a spectral Galerkin solver. Across eight benchmark fluid and plasma systems, including high-dimensional turbulent simulations, WARP produces numerically stable rollouts from new random initial conditions and substantially improves coefficient recovery, rollout accuracy, and statistical measures compared with a single-step weak-form SINDy baseline.
ER -