Learning to Emulate Chaos: Adversarial Optimal Transport Regularization

arXiv:2604.21097v1 Announce Type: cross Abstract: Chaos arises in many complex dynamical systems, from weather to power grids, but is difficult to accurately model using data-driven emulators, including neural operator architectures. For chaotic systems, the inherent sensitivity to initial conditions makes exact long-term forecasts theoretically infeasible, meaning that traditional squared-error losses often fail when trained on noisy data. Recent work has focused on training emulators to match the statistical properties of chaotic attractors by introducing regularization based on handcrafted local features and summary statistics, as well as learned statistics extracted from a diverse dataset of trajectories. In this work, we propose a family of adversarial optimal transport objectives that jointly learn high-quality summary statistics and a physically consistent emulator. We theoretically analyze and experimentally validate a Sinkhorn divergence formulation (2-Wasserstein) and a WGAN-style dual formulation (1-Wasserstein). Our experiments across a variety of chaotic systems, including systems with high-dimensional chaotic attractors, show that emulators trained with our approach exhibit significantly improved long-term statistical fidelity.

Related

• Data-driven Reachable Set Estimation with Tunable Adversarial and Wasserstein Distributional Guarantees
• Nonparametric Sparse Online Learning of the Koopman Operator
• Transformers for dynamical systems learn transfer operators in-context
• A robust and adaptive MPC formulation for Gaussian process models
• Amortized Filtering and Smoothing with Conditional Normalizing Flows

Source: arXiv cs.LG | 2026-04-24