Addressing A Posteriori Performance Degradation in Neural Network Subgrid Stress Models

arXiv:2511.17475 · physics.flu-dyn, cs.LG · Submitted 2025-11-21 · Read on arXiv

physics.flu-dyn, cs.LG

Submitted: 2025-11-21

Updated: 2026-09-10

DOI: 10.1017/jfm.2026.11995

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Distribution Shift — 10× in this paper · explained in 5 episodes
• Data Augmentation — 8× in this paper · explained in 7 episodes
• Robustness — 7× in this paper · explained in 5 episodes
• Artificial Neural Network — 4× in this paper
• Large Eddy Simulations — 4× in this paper
• Reinforcement Learning — 4× in this paper · explained in 4 episodes
• Ablation — 2× in this paper · explained in 1 episode
• Deep Reinforcement Learning — 2× in this paper · explained in 2 episodes
• Trace — 2× in this paper · explained in 2 episodes
• Accuracy — 1× in this paper · explained in 1 episode
• Activation Function — 1× in this paper · explained in 1 episode
• Direct Numerical Simulation — 1× in this paper
• Fidelity — 1× in this paper · explained in 2 episodes
• Generalization — 1× in this paper · explained in 13 episodes
• Graph Neural Network (GNN) — 1× in this paper · explained in 2 episodes
• Loss Function — 1× in this paper · explained in 2 episodes
• Transferability — 1× in this paper · explained in 4 episodes

Sources

• Stable Machine-Learning Parameterization of Subgrid Processes in a Comprehensive Atmospheric Model Learned From Embedded Convection-Permitting Simulations
• Sensitivity and Generalization in Neural Networks: an Empirical Study
• Subgrid Stress Modelling with Multi-dimensional State Space Sequence Models

Related papers

• Reward hacking in physical reinforcement learning revealed by turbulent drag reduction
• Diffraction of walking drops by a standing Faraday wave
• Adjoint-based shape optimization of a ship hull using a Conditional Variational Autoencoder (CVAE) assisted propulsion surrogate model