Embedding physical symmetries into machine-learned reduced plasma physics models via data augmentation
arXiv:2506.14048 · doi:10.1103/9fq1-vpqz
Abstract
Machine learning is offering powerful new tools for the development and discovery of reduced models of nonlinear, multiscale plasma dynamics from the data of first-principles kinetic simulations. However, ensuring the physical consistency of such models requires embedding fundamental symmetries of plasma dynamics. In this work, we explore a symmetry-embedding strategy based on data augmentation, where symmetry-preserving transformations (e.g., Lorentz and Galilean boosts) are applied to simulation data. Using both sparse regression and neural networks, we show that models trained on symmetry-augmented data more accurately infer the plasma fluid equations and pressure tensor closures from fully kinetic particle-in-cell simulations of magnetic reconnection. We show that this approach suppresses spurious inertial-frame-dependent correlations between dynamical variables, improves data efficiency, and significantly outperforms models trained without symmetry-augmented data, as well as commonly used theoretical pressure closure models. Our results establish symmetry-based data augmentation as a broadly applicable method for incorporating physical structure into machine-learned reduced plasma models.
Corrected Figure 1
References in corpus (29)
- Discovering governing equations from data: Sparse identification of nonlinear dynamical systems
- Deep Big Simple Neural Nets Excel on Handwritten Digit Recognition
- Relativistic Reconnection: an Efficient Source of Non-Thermal Particles
- Predicting catastrophes in nonlinear dynamical systems by compressive sensing
- SINDy-PI: A Robust Algorithm for Parallel Implicit Sparse Identification of Nonlinear Dynamics
- Weak SINDy For Partial Differential Equations
- Particle Acceleration in Relativistic Plasma Turbulence
- One-to-one direct modeling of experiments and astrophysical scenarios: pushing the envelope on kinetic plasma simulations
- A Review of the 0.1 Reconnection Rate Problem
- Using Noisy or Incomplete Data to Discover Models of Spatiotemporal Dynamics
- Physics-constrained, low-dimensional models for MHD: First-principles and data-driven approaches
- Efficient Nonthermal Particle Acceleration by the Kink Instability in Relativistic Jets
- First-Principles Theory of the Rate of Magnetic Reconnection in Magnetospheric and Solar Plasmas
- Data-driven discovery of reduced plasma physics models from fully-kinetic simulations
- Machine Learning Surrogate Models for Landau Fluid Closure
- Machine learning moment closure models for the radiative transfer equation I: directly learning a gradient based closure
- Neural network representability of fully ionized plasma fluid model closures
- Permanent magnet optimization for stellarators as sparse regression
- Data-driven modeling of Landau damping by physics-informed neural networks
- Learning fluid physics from highly turbulent data using sparse physics-informed discovery of empirical relations (SPIDER)
- Effective resistivity in relativistic collisionless plasmoid-mediated reconnection
- Machine-learning heat flux closure for multi-moment fluid modeling of nonlinear Landau damping
- In Search of a Data Driven Symbolic Multi-Fluid 10-Moment Model Closure
- Identification of high order closure terms from fully kinetic simulations using machine learning
- Data-driven discovery of a heat flux closure for electrostatic plasma phenomena
- Relativistic reconnection with effective resistivity: I. Dynamics and reconnection rate
- Data-driven sparse modeling of oscillations in plasma space propulsion
- Learning the dynamics of a one-dimensional plasma model with graph neural networks
- Machine-learning Closure for Vlasov-Poisson Dynamics in Fourier-Hermite Space