Effective models and predictability of chaotic multiscale systems via machine learning

Abstract

We scrutinize the use of machine learning, based on reservoir computing, to build data-driven effective models of multiscale chaotic systems. We show that, for a wide scale separation, machine learning generates effective models akin to those obtained using multiscale asymptotic techniques and, remarkably, remains effective in predictability also when the scale separation is reduced. We also show that predictability can be improved by hybridizing the reservoir with an imperfect model.

arXiv comment12 pages with 10 figures. Accepted in Physical Review E

References in corpus (16)

• Machine learning and the physical sciences
• Recent Advances in Physical Reservoir Computing: A Review
• Using Machine Learning to Replicate Chaotic Attractors and Calculate Lyapunov Exponents from Data
• Attractor Reconstruction by Machine Learning
• Data-Driven Forecasting of High-Dimensional Chaotic Systems with Long Short-Term Memory Networks
• Hybrid Forecasting of Chaotic Processes: Using Machine Learning in Conjunction with a Knowledge-Based Model
• Data-assisted reduced-order modeling of extreme events in complex dynamical systems
• Growth of non-infinitesimal perturbations in turbulence
• An extension of the Lyapunov analysis for the predictability problem
• Machine-learning inference of fluid variables from data using reservoir computing
• Local Dynamics in Trained Recurrent Neural Networks
• Combining Machine Learning with Knowledge-Based Modeling for Scalable Forecasting and Subgrid-Scale Closure of Large, Complex, Spatiotemporal Systems
• Optimal subgrid scheme for shell models of turbulence
• The predictability problem in systems with an uncertainty in the evolution law
• Lyapunov analysis of multiscale dynamics: The slow bundle of the two-scale Lorenz 96 model
• Physics-Based Learning Models for Ship Hydrodynamics

Cited by in corpus (8)

• Reservoir Computing with Diverse Timescales for Prediction of Multiscale Dynamics
• Using Machine Learning to Anticipate Tipping Points and Extrapolate to Post-Tipping Dynamics of Non-Stationary Dynamical Systems
• Model-free prediction of multistability using echo state network
• Stability analysis of chaotic systems from data
• Predicting multi-parametric dynamics of externally forced oscillator using reservoir computing and minimal data
• Curriculum learning for data-driven modeling of dynamical systems
• Role of assortativity in predicting burst synchronization using echo state network
• Using machine-learning modelling to understand macroscopic dynamics in a system of coupled maps