EM-like Learning Chaotic Dynamics from Noisy and Partial Observations

Abstract

The identification of the governing equations of chaotic dynamical systems from data has recently emerged as a hot topic. While the seminal work by Brunton et al. reported proof-of-concepts for idealized observation setting for fully-observed systems, {\em i.e.} large signal-to-noise ratios and high-frequency sampling of all system variables, we here address the learning of data-driven representations of chaotic dynamics for partially-observed systems, including significant noise patterns and possibly lower and irregular sampling setting. Instead of considering training losses based on short-term prediction error like state-of-the-art learning-based schemes, we adopt a Bayesian formulation and state this issue as a data assimilation problem with unknown model parameters. To solve for the joint inference of the hidden dynamics and of model parameters, we combine neural-network representations and state-of-the-art assimilation schemes. Using iterative Expectation-Maximization (EM)-like procedures, the key feature of the proposed inference schemes is the derivation of the posterior of the hidden dynamics. Using a neural-network-based Ordinary Differential Equation (ODE) representation of these dynamics, we investigate two strategies: their combination to Ensemble Kalman Smoothers and Long Short-Term Memory (LSTM)-based variational approximations of the posterior. Through numerical experiments on the Lorenz-63 system with different noise and time sampling settings, we demonstrate the ability of the proposed schemes to recover and reproduce the hidden chaotic dynamics, including their Lyapunov characteristic exponents, when classic machine learning approaches fail.

References in corpus (9)

• Discovering governing equations from data: Sparse identification of nonlinear dynamical systems
• A Recurrent Latent Variable Model for Sequential Data
• Using Machine Learning to Replicate Chaotic Attractors and Calculate Lyapunov Exponents from Data
• Variational Autoencoder for Deep Learning of Images, Labels and Captions
• Data Driven Governing Equations Approximation Using Deep Neural Networks
• Neural Variational Inference and Learning in Belief Networks
• Multistep Neural Networks for Data-driven Discovery of Nonlinear Dynamical Systems
• Deep learning algorithm for data-driven simulation of noisy dynamical system
• Bilinear residual Neural Network for the identification and forecasting of dynamical systems

Cited by in corpus (9)

• Data-Driven Deep Learning of Partial Differential Equations in Modal Space
• Supervised learning from noisy observations: Combining machine-learning techniques with data assimilation
• On generalized residue network for deep learning of unknown dynamical systems
• Using Machine Learning to Anticipate Tipping Points and Extrapolate to Post-Tipping Dynamics of Non-Stationary Dynamical Systems
• Learning Latent Dynamics for Partially-Observed Chaotic Systems
• Symplectic Learning for Hamiltonian Neural Networks
• Combining data assimilation and machine learning to infer unresolved scale parametrisation
• Variational Deep Learning for the Identification and Reconstruction of Chaotic and Stochastic Dynamical Systems from Noisy and Partial Observations
• Machine Learning Methods for Autonomous Ordinary Differential Equations