Analog forecasting with dynamics-adapted kernels
From MaRDI portal
Abstract: Analog forecasting is a nonparametric technique introduced by Lorenz in 1969 which predicts the evolution of states of a dynamical system (or observables defined on the states) by following the evolution of the sample in a historical record of observations which most closely resembles the current initial data. Here, we introduce a suite of forecasting methods which improve traditional analog forecasting by combining ideas from kernel methods developed in harmonic analysis and machine learning and state-space reconstruction for dynamical systems. A key ingredient of our approach is to replace single-analog forecasting with weighted ensembles of analogs constructed using local similarity kernels. The kernels used here employ a number of dynamics-dependent features designed to improve forecast skill, including Takens' delay-coordinate maps (to recover information in the initial data lost through partial observations) and a directional dependence on the dynamical vector field generating the data. Mathematically, our approach is closely related to kernel methods for out-of-sample extension of functions, and we discuss alternative strategies based on the Nystr"om method and the multiscale Laplacian pyramids technique. We illustrate these techniques in applications to forecasting in a low-order deterministic model for atmospheric dynamics with chaotic metastability, and interannual-scale forecasting in the North Pacific sector of a comprehensive climate model. We find that forecasts based on kernel-weighted ensembles have significantly higher skill than the conventional approach following a single analog.
Recommendations
- Kernel Analog Forecasting: Multiscale Test Problems
- Operator-theoretic framework for forecasting nonlinear time series with kernel analog techniques
- Prediction of dynamical time series using kernel based regression and smooth splines
- A model‐based approach for analog spatio‐temporal dynamic forecasting
- Kernel-based prediction of non-Markovian time series
- Combining Forecasts with Nonparametric Kernel Regressions
- Prediction With Approximated Gaussian Process Dynamical Models
- Adaptive prediction in the presence of unmodelled dynamics
Cited in
(25)- Reduced-space Gaussian process regression for data-driven probabilistic forecast of chaotic dynamical systems
- Kernel-based prediction of non-Markovian time series
- Operator-theoretic framework for forecasting nonlinear time series with kernel analog techniques
- Machine learning techniques to construct patched analog ensembles for data assimilation
- Discrepancy-based theory and algorithms for forecasting non-stationary time series
- Projecting flood-inducing precipitation with a Bayesian analogue model
- Data-driven spectral decomposition and forecasting of ergodic dynamical systems
- Spatiotemporal pattern extraction by spectral analysis of vector-valued observables
- Nonlinear analog predictor analysis: A coupled neural network/analog model for climate downscaling
- The diffusion geometry of fibre bundles: horizontal diffusion maps
- Data-driven forecasting of high-dimensional chaotic systems with long short-term memory networks
- Koopman analysis of the long-term evolution in a turbulent convection cell
- Kernel Analog Forecasting: Multiscale Test Problems
- Generative stochastic modeling of strongly nonlinear flows with non-Gaussian statistics
- Higher order dynamic mode decomposition
- Lie group valued Koopman eigenfunctions
- The mpEDMD Algorithm for Data-Driven Computations of Measure-Preserving Dynamical Systems
- Learning Theory for Dynamical Systems
- Learning to Forecast Dynamical Systems from Streaming Data
- Rigorous data‐driven computation of spectral properties of Koopman operators for dynamical systems
- Spatiotemporal analysis using Riemannian composition of diffusion operators
- Ensemble forecasts in reproducing kernel Hilbert space family
- Conditional expectation using compactification operators
- Learning nonparametric ordinary differential equations from noisy data
- Deep echo state networks with uncertainty quantification for spatio-temporal forecasting
This page was built for publication: Analog forecasting with dynamics-adapted kernels
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2821966)
