On: Model scale vs. domain knowledge in statistical forecasting of chaotic systems

[westurner]

"Can Machine Learning Predict Chaos? This Paper from UT Austin Performs a Large-Scale Comparison of Modern Forecasting Methods on a Giant Dataset of 135 Chaotic Systems" https://www.marktechpost.com/2023/12/25/can-machine-learning... :

From https://twitter.com/wgilpin0/status/1737934964757565848 :

> I found that large domain-agnostic models (Transformers, LSTM, etc) can forecast chaos really far into the future (>10 Lyapunov times). With enough training history, they outperform physics methods (next-gen reservoir computers, neural ODE, etc).

[westurner → westurner]

Is there any advantage to playing fluid [wavefield,] recordings in reverse with their parameters as training data?