Operator-theoretic framework for forecasting nonlinear time series with kernel analog techniques

DOI10.1016/J.PHYSD.2020.132520zbMATH Open1496.37085arXiv1906.00464OpenAlexW3016578580MaRDI QIDQ2125604FDOQ2125604

Authors: Romeo Alexander, Dimitrios Giannakis 

Publication date: 14 April 2022

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

Abstract: Kernel analog forecasting (KAF), alternatively known as kernel principal component regression, is a kernel method used for nonparametric statistical forecasting of dynamically generated time series data. This paper synthesizes descriptions of kernel methods and Koopman operator theory in order to provide a single consistent account of KAF. The framework presented here illuminates the property of the KAF method that, under measure-preserving and ergodic dynamics, it consistently approximates the conditional expectation of observables that are acted upon by the Koopman operator of the dynamical system and are conditioned on the observed data at forecast initialization. More precisely, KAF yields optimal predictions, in the sense of minimal root mean square error with respect to the invariant measure, in the asymptotic limit of large data. The presented framework facilitates, moreover, the analysis of generalization error and quantification of uncertainty. Extensions of KAF to the construction of conditional variance and conditional probability functions, as well as to non-symmetric kernels, are also shown. Illustrations of various aspects of KAF are provided with applications to simple examples, namely a periodic flow on the circle and the chaotic Lorenz 63 system.

Full work available at URL: https://arxiv.org/abs/1906.00464

Recommendations

• Kernel-based prediction of non-Markovian time series
• Analog forecasting with dynamics-adapted kernels
• Kernel Analog Forecasting: Multiscale Test Problems
• Prediction of high-dimensional time series with exogenous variables using generalized Koopman operator framework in reproducing kernel Hilbert space
• Nonstationary time series forecasting using wavelets and kernel smoothing
• Koopman operator framework for time series modeling and analysis
• A Functional Wavelet–Kernel Approach for Time Series Prediction
• Identification of non-linear time series via kernels

zbMATH Keywords

kernel methodsconditional expectationstatistical forecastingKoopman operators

Mathematics Subject Classification ID

Time series analysis of dynamical systems (37M10)

Cites Work

• Diffusion maps
• On Learning Vector-Valued Functions
• Theory of Reproducing Kernels
• Learning Theory
• Support Vector Machines
• Foundations of Modern Probability
• Universality, Characteristic Kernels and RKHS Embedding of Measures
• Kernel methods in machine learning
• Consistency of spectral clustering
• Dynamic mode decomposition of numerical and experimental data
• Spectral analysis of nonlinear flows
• Deterministic Nonperiodic Flow
• Embedology
• What are SRB measures, and which dynamical systems have them?
• Theoretical foundations of the potential function method in pattern recognition learning
• 10.1162/15324430260185646
• Title not available (Why is that?)
• Title not available (Why is that?)
• Laplacian Eigenmaps for Dimensionality Reduction and Data Representation
• From graph to manifold Laplacian: the convergence rate
• Local kernels and the geometric structure of data
• Time-scale separation from diffusion-mapped delay coordinates
• On the Approximation of Complicated Dynamical Behavior
• The Lorenz attractor exists
• Turbulence, Coherent Structures, Dynamical Systems and Symmetry
• Learning Theory
• Interpolation of scattered data: distance matrices and conditionally positive definite functions
• Variable bandwidth diffusion kernels
• Eigenvalue distribution of compact operators
• Title not available (Why is that?)
• Operator theoretic aspects of ergodic theory
• Nonlinear prediction of chaotic time series
• Bi-stochastic kernels via asymmetric affinity functions
• Geometric harmonics: a novel tool for multiscale out-of-sample extension of empirical functions
• On learning with integral operators
• Traceable integral kernels on countably generated measure spaces
• Integral manifolds and inertial manifolds for dissipative partial differential equations
• Eigenvalues of integral operators defined by smooth positive definite kernels
• Comparison of systems with complex behavior
• Spectral properties of dynamical systems, model reduction and decompositions
• Applied Koopmanism
• Title not available (Why is that?)
• A data-driven approximation of the koopman operator: extending dynamic mode decomposition
• Extracting qualitative dynamics from experimental data
• A computational method to extract macroscopic variables and their dynamics in multiscale systems
• On the numerical approximation of the Perron-Frobenius and Koopman operator
• Estimating Long-Term Behavior of Flows without Trajectory Integration: The Infinitesimal Generator Approach
• A topological delay embedding theorem for infinite-dimensional dynamical systems
• Hamiltonian Systems and Transformation in Hilbert Space
• The Lorenz attractor is mixing
• Analysis of the 3DVAR filter for the partially observed Lorenz '63 model
• Eigenvalue decay rates for positive integral operators
• Title not available (Why is that?)
• An introduction to the theory of reproducing kernel Hilbert spaces
• Singular spectrum analysis in nonlinear dynamics, with applications to paleoclimatic time series
• Diffusion forecasting model with basis functions from QR-decomposition
• Data-driven model reduction and transfer operator approximation
• Nonlinear Laplacian spectral analysis for time series with intermittency and low-frequency variability
• Positive definiteness, reproducing kernel Hilbert spaces and beyond
• Data-driven spectral analysis of the Koopman operator
• Delay-coordinate maps and the spectra of Koopman operators
• Data-driven spectral decomposition and forecasting of ergodic dynamical systems
• Ergodic theory, dynamic mode decomposition, and computation of spectral properties of the Koopman operator
• Comparisons of new nonlinear modeling techniques with applications to infant respiration.
• Title not available (Why is that?)
• Analog forecasting with dynamics-adapted kernels
• Eigendecompositions of transfer operators in reproducing kernel Hilbert spaces
• Kernel flows: from learning kernels from data into the abyss
• Spatiotemporal pattern extraction by spectral analysis of vector-valued observables

Cited In (23)

• Learning dynamical systems from data: a simple cross-validation perspective. III: Irregularly-sampled time series
• Simplicity bias, algorithmic probability, and the random logistic map
• Bridging algorithmic information theory and machine learning: a new approach to kernel learning
• Hausdorff metric based training of kernels to learn attractors with application to 133 chaotic dynamical systems
• Analog forecasting with dynamics-adapted kernels
• One-shot learning of stochastic differential equations with data adapted kernels
• Koopman operator framework for time series modeling and analysis
• A framework for machine learning of model error in dynamical systems
• Reproducing kernel Hilbert space compactification of unitary evolution groups
• Learning Theory for Dynamical Systems
• Prediction of high-dimensional time series with exogenous variables using generalized Koopman operator framework in reproducing kernel Hilbert space
• Kernel Analog Forecasting: Multiscale Test Problems
• Nonlinear model reduction for slow-fast stochastic systems near unknown invariant manifolds
• Ensemble forecasts in reproducing kernel Hilbert space family
• Kernel-based prediction of non-Markovian time series
• Learning dynamical systems from data: a simple cross-validation perspective. IV: Case with partial observations
• Learning to Forecast Dynamical Systems from Streaming Data
• Error bounds of the invariant statistics in machine learning of ergodic Itô diffusions
• Learning dynamical systems from data: a simple cross-validation perspective. V: Sparse kernel flows for 132 chaotic dynamical systems
• Solving PDEs on unknown manifolds with machine learning
• Kernel methods for center manifold approximation and a weak data-based version of the center manifold theorem
• Learning dynamical systems from data: a simple cross-validation perspective. I: Parametric kernel flows
• A note on microlocal kernel design for some slow-fast stochastic differential equations with critical transitions and application to EEG signals

This page was built for publication: Operator-theoretic framework for forecasting nonlinear time series with kernel analog techniques

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2125604)