Causally constrained reduced-order neural models of complex turbulent dynamical systems

2026-02-14

AI Summary

This work addresses the limited response capability to external forcing in reduced-order neural models of complex turbulent systems, which often arises from spurious non-causal dependencies. To overcome this, the authors propose a modeling framework that integrates explicit causal constraints grounded in response theory and score matching. Notably, the approach enforces causality using only unforced training data, effectively suppressing non-physical correlations. Experiments on the stochastic Charney–DeVore model demonstrate that the proposed method substantially enhances the model’s generalization to external forcings—ranging from weak to strong—as well as its predictive robustness and physical consistency. This study presents the first successful integration of causal constraints into reduced-order neural modeling, offering a general and scalable paradigm for modeling turbulent systems such as those encountered in climate dynamics.

Technology Category

• Machine Learning: Causal Learning
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Reasoning under Uncertainty: Causality

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

We introduce a flexible framework based on response theory and score matching to suppress spurious, noncausal dependencies in reduced-order neural emulators of turbulent systems, focusing on climate dynamics as a proof-of-concept. We showcase the approach using the stochastic Charney-DeVore model as a relevant prototype for low-frequency atmospheric variability. We show that the resulting causal constraints enhance neural emulators'ability to respond to both weak and strong external forcings, despite being trained exclusively on unforced data. The approach is broadly applicable to modeling complex turbulent dynamical systems in reduced spaces and can be readily integrated into general neural network architectures.

Problem

• causal constraints
• reduced-order modeling
• turbulent dynamical systems
• neural emulators
• spurious dependencies

Innovation

• causal constraints
• reduced-order modeling
• score matching
• response theory
• turbulent dynamical systems