Echo state networks trained by Tikhonov least squares are \(L^2(\mu)\) approximators of ergodic dynamical systems

DOI10.1016/j.physd.2021.132882zbMath1491.37075arXiv2005.06967WikidataQ114141965 ScholiaQ114141965MaRDI QIDQ2077652

Allen Hart, James Hook, Jonathan H. P. Dawes

Publication date: 21 February 2022

Published in: Physica D (Search for Journal in Brave)

Full work available at URL: https://arxiv.org/abs/2005.06967

zbMATH Keywords

Lorenz equations; time series analysis; recurrent neural networks; reservoir computing; liquid state machine; delay embedding

Mathematics Subject Classification ID

37M10: Time series analysis of dynamical systems

37M25: Computational methods for ergodic theory (approximation of invariant measures, computation of Lyapunov exponents, entropy, etc.)

Related Items

Symmetry kills the square in a multifunctional reservoir computer, Reservoir Computing with an Inertial Form, Dimension reduction in recurrent networks by canonicalization

Uses Software

• RODES

Cites Work

• Unnamed Item
• Unnamed Item
• Unnamed Item
• Unnamed Item
• The Lorenz attractor is mixing
• Structural stability of Lorenz attractors
• The structure of Lorenz attractors
• Mathematical problems for the next century
• A rigorous ODE solver and Smale's 14th problem
• Re-visiting the echo state property
• Data-driven reconstruction of nonlinear dynamics from sparse observation
• Automatic speech recognition using a predictive echo state network classifier
• Learning grammatical structure with Echo State Networks
• Deblurring Images
• Dynamical systems with generalized hyperbolic attractors: hyperbolic, ergodic and topological properties
• The Lorenz attractor exists
• The rate of convergence in ergodic theorems
• Real-Time Computing Without Stable States: A New Framework for Neural Computation Based on Perturbations
• Using machine learning to replicate chaotic attractors and calculate Lyapunov exponents from data