Learning Theory for Dynamical Systems

DOI10.1137/22M1516865MaRDI QIDQ6132792zbMATH OpenOpenAlexFDO

Authors T. Berry, Suddhasattwa Das

Publication date 17 August 2023

Published in SIAM Journal on Applied Dynamical Systems (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2208.05349

zbMATH Keywords

mixingLyapunov exponentreservoir computingmatrix cocycledirect forecastdelay-coordinatesiterative forecast

Mathematics Subject Classification ID

Time series analysis of dynamical systems (37M10) Uniformly hyperbolic systems (expanding, Anosov, Axiom A, etc.) (37D20) Nonuniformly hyperbolic systems (Lyapunov exponents, Pesin theory, etc.) (37D25) Simulation of dynamical systems (37M05) Computational methods for ergodic theory (approximation of invariant measures, computation of Lyapunov exponents, entropy, etc.) (37M25) Dynamical systems in numerical analysis (37N30)

Abstract: The task of modelling and forecasting a dynamical system is one of the oldest problems, and it remains challenging. Broadly, this task has two subtasks - extracting the full dynamical information from a partial observation; and then explicitly learning the dynamics from this information. We present a mathematical framework in which the dynamical information is represented in the form of an embedding. The framework combines the two subtasks using the language of spaces, maps, and commutations. The framework also unifies two of the most common learning paradigms - delay-coordinates and reservoir computing. We use this framework as a platform for two other investigations of the reconstructed system - its dynamical stability; and the growth of error under iterations. We show that these questions are deeply tied to more fundamental properties of the underlying system - the behavior of matrix cocycles over the base dynamics, its non-uniform hyperbolic behavior, and its decay of correlations. Thus, our framework bridges the gap between universally observed behavior of dynamics modelling; and the spectral, differential and ergodic properties intrinsic to the dynamics.

Recommendations

• Learning latent dynamics for partially observed chaotic systems
• Learning strange attractors with reservoir systems
• Learning dynamics by reservoir computing (In Memory of Prof. Pavol Brunovský)
• Koopman operator, geometry, and learning of dynamical systems

Cites work

• scientific article; zbMATH DE number 3121911 (Why is no real title available?)
• scientific article; zbMATH DE number 19532 (Why is no real title available?)
• (Dis)continuity of Lyapunov exponents
• A concept of homeomorphic defect for defining mostly conjugate dynamical systems
• A variational approach to the consistency of spectral clustering
• Analog forecasting with dynamics-adapted kernels
• Approximation of Bernoulli measures for non-uniformly hyperbolic systems
• Coherent structures and isolated spectrum for Perron–Frobenius cocycles
• Deep learning of conjugate mappings
• Delay-coordinate maps and the spectra of Koopman operators
• Diffusion maps
• Echo state networks trained by Tikhonov least squares are \(L^2(\mu)\) approximators of ergodic dynamical systems
• Embedding and approximation theorems for echo state networks
• Embedology
• Equivalence of physical and SRB measures in random dynamical systems
• Ergodic theory of differentiable dynamical systems
• Erratum: “On explaining the surprising success of reservoir computing forecaster of chaos? The universal machine learning dynamical system with contrasts to VAR and DMD” [Chaos 31(1), 013108 (2021)]
• FAMILIES OF INVARIANT MANIFOLDS CORRESPONDING TO NONZERO CHARACTERISTIC EXPONENTS
• Fading memory echo state networks are universal
• Identification and prediction of low dimensional dynamics
• Is the largest Lyapunov exponent preserved in embedded dynamics?
• Koopman spectra in reproducing kernel Hilbert spaces
• Learning strange attractors with reservoir systems
• Liapunov multipliers and decay of correlations in dynamical systems
• Lyapunov exponents and rates of mixing for one-dimensional maps
• Lyapunov exponents of hyperbolic measures and hyperbolic periodic orbits
• Lyapunov exponents of linear cocycles. Continuity via large deviations
• Lyapunov exponents, entropy and periodic orbits for diffeomorphisms
• Machine learning for prediction with missing dynamics
• Markov structures and decay of correlations for non-uniformly expanding dynamical systems
• Model reduction with memory and the machine learning of dynamical systems
• Nonlinear Regression
• Nonlinear prediction of chaotic time series
• Numerical orbits of chaotic processes represent true orbits
• Operator-theoretic framework for forecasting nonlinear time series with kernel analog techniques
• Optimal Construction of Koopman Eigenfunctions for Prediction and Control
• Products of Random Matrices
• Quantitative Pesin theory for Anosov diffeomorphisms and flows
• Random matrix products and measures on projective spaces
• Recurrence times and rates of mixing
• Regularity of invariant graphs for forced systems
• Regularized local linear prediction of chaotic time series
• Reproducing kernel Hilbert space compactification of unitary evolution groups
• Reservoir computing approaches to recurrent neural network training
• SRB measures as zero-noise limits
• Smoothness of invariant manifolds for nonautonomous equations
• Stable manifolds for nonautonomous equations without exponential dichotomy
• Statistical properties of dynamical systems with some hyperbolicity
• Stochastic nonlinear time series forecasting using time-delay reservoir computers: performance and universality
• Super convergence of ergodic averages for quasiperiodic orbits
• The Lyapunov exponents of generic volume-preserving and symplectic maps
• The metric entropy of diffeomorphisms. I: Characterization of measures satisfying Pesin's entropy formula
• The metric entropy of diffeomorphisms. II: Relations between entropy, exponents and dimension
• The topological invariance of Lyapunov exponents in embedded dynamics
• Time series analysis. Forecasting and control
• Time-scale separation from diffusion-mapped delay coordinates
• Time-series learning of latent-space dynamics for reduced-order model closure
• Using machine learning to replicate chaotic attractors and calculate Lyapunov exponents from data

Cited in

(8)

• Data-driven cold starting of good reservoirs
• scientific article; zbMATH DE number 65825 (Why is no real title available?)
• Learning about dynamical systems via unfalsification of hypotheses
• Learning strange attractors with reservoir systems
• A proposal on machine learning via dynamical systems
• Riddled basins of the optimal states in learning dynamical systems
• LEARNING TRAJECTORIES OF DYNAMICAL SYSTEMS
• Fluctuation-Dissipation Theorem and Models of Learning

This page was built for publication: Learning Theory for Dynamical Systems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6132792)