Abstract:Gaussian processes provide a flexible, non-parametric framework for the approximation of functions in high-dimensional spaces. The covariance kernel is the main engine of Gaussian processes, incorporating correlations that underpin the predictive distribution. For applications with spatiotemporal datasets, suitable kernels should model joint spatial and temporal dependence. Separable space-time covariance kernels offer simplicity and computational efficiency. However, non-separable kernels include space-time interactions that better capture observed correlations. Most non-separable kernels that admit explicit expressions are based on mathematical considerations (admissibility conditions) rather than first-principles derivations. We present a hybrid spectral approach for generating covariance kernels which is based on physical arguments. We use this approach to derive a new class of physically motivated, non-separable covariance kernels which have their roots in the stochastic, linear, damped, harmonic oscillator (LDHO). The new kernels incorporate functions with both monotonic and oscillatory decay of space-time correlations. The LDHO covariance kernels involve space-time interactions which are introduced by dispersion relations that modulate the oscillator coefficients. We derive explicit relations for the spatiotemporal covariance kernels in the three oscillator regimes (underdamping, critical damping, overdamping) and investigate their properties. We further illustrate the hybrid spectral method by deriving covariance kernels that are based on the Ornstein-Uhlenbeck model.

Nonstationary covariance functions that model space–time interactions

Spatio-Temporal Stationary Covariance Models

Spatio-temporal Variograms and Covariance Models

A Nonstationary Designer Space-Time Kernel

Families of spatio-temporal stationary covariance models

Spatio-Temporal Covariance Functions Generated by Mixtures

Nonstationary Spatial Process Models with Spatially Varying Covariance Kernels

Non-separable Covariance Kernels for Spatiotemporal Gaussian Processes based on a Hybrid Spectral Method and the Harmonic Oscillator

Intrinsically Stationary Variograms in Space and Time

Recent developments on the construction of spatio-temporal covariance models

Linear Combinations of Space-Time Covariance Functions and Variograms

The Use of the Variogram in Construction of Stationary Time Series Models

Spatiotemporal Covariance Neural Networks

Semiparametric Spatio-Temporal Covariance Models with the ARMA Temporal Margin

Modeling nonstationarity in space and time

Planar-Temporal Stationary Correlation Models That Depend on the Maximum Norm

Hybrid Parametric Classes of Isotropic Covariance Functions for Spatial Random Fields

Modelling Function-Valued Processes with Nonseparable Covariance Structure

Space-Time Models based on Random Fields with Local Interactions

Modeling Function-Valued Processes with Nonseparable and/or Nonstationary Covariance Structure.

Compatibility of Space-Time Kernels with Full, Dynamical, or Compact Support