Scalable computations for nonstationary Gaussian processes
From MaRDI portal
Publication:6173565
DOI10.1007/S11222-023-10252-0zbMATH Open1516.62005arXiv2206.05220OpenAlexW4381281560MaRDI QIDQ6173565FDOQ6173565
Authors: Paul G. Beckman, Christopher J. Geoga, Michael L. Stein, Mihai Anitescu 
Publication date: 21 July 2023
Published in: Statistics and Computing (Search for Journal in Brave)
Abstract: Nonstationary Gaussian process models can capture complex spatially varying dependence structures in spatial datasets. However, the large number of observations in modern datasets makes fitting such models computationally intractable with conventional dense linear algebra. In addition, derivative-free or even first-order optimization methods can be slow to converge when estimating many spatially varying parameters. We present here a computational framework that couples an algebraic block-diagonal plus low-rank covariance matrix approximation with stochastic trace estimation to facilitate the efficient use of second-order solvers for maximum likelihood estimation of Gaussian process models with many parameters. We demonstrate the effectiveness of these methods by simultaneously fitting 192 parameters in the popular nonstationary model of Paciorek and Schervish using 107,600 sea surface temperature anomaly measurements.
Full work available at URL: https://arxiv.org/abs/2206.05220
Recommendations
- A Full Scale Approximation of Covariance Functions for Large Spatial Data Sets
- Bayesian nonstationary Gaussian process models via treed process convolutions
- Efficient estimation of nonstationary spatial covariance functions with application to high-resolution climate model emulation
- Kryging: geostatistical analysis of large-scale datasets using Krylov subspace methods
- Smoothed full-scale approximation of Gaussian process models for computation of large spatial data sets
Computational methods for problems pertaining to statistics (62-08) Gaussian processes (60G15) Applications of statistics to environmental and related topics (62P12)
Cites Work
- Gaussian process learning via Fisher scoring of Vecchia's approximation
- Fixed Rank Kriging for Very Large Spatial Data Sets
- Numerical Optimization
- Interpolation of spatial data. Some theory for kriging
- Gaussian Predictive Process Models for Large Spatial Data Sets
- Approximating Likelihoods for Large Spatial Data Sets
- Inconsistent Estimation and Asymptotically Equal Interpolations in Model-Based Geostatistics
- A general framework for Vecchia approximations of Gaussian processes
- Local likelihood estimation for nonstationary random fields
- An explicit link between Gaussian fields and Gaussian Markov random fields: the stochastic partial differential equation approach
- A Full Scale Approximation of Covariance Functions for Large Spatial Data Sets
- A class of multi-resolution approximations for large spatial datasets
- Hilbert space methods for reduced-rank Gaussian process regression
- Covariance approximation for large multivariate spatial data sets with an application to multiple climate model errors
- Scalable Gaussian Process Computations Using Hierarchical Matrices
- Approximate Bayesian inference for large spatial datasets using predictive process models
- A Stochastic Estimator of the Trace of the Influence Matrix for Laplacian Smoothing Splines
- Stochastic approximation of score functions for Gaussian processes
- Multidimensional binary search trees used for associative searching
- Hierarchically compositional kernels for scalable nonparametric learning
- Fast spatial Gaussian process maximum likelihood estimation via skeletonization factorizations
- Likelihood approximation with hierarchical matrices for large spatial datasets
- Fitting Matérn smoothness parameters using automatic differentiation
- Efficient estimation of nonstationary spatial covariance functions with application to high-resolution climate model emulation
Cited In (15)
- Efficient estimation of nonstationary spatial covariance functions with application to high-resolution climate model emulation
- Bayesian nonstationary Gaussian process models via treed process convolutions
- A scalable gaussian process analysis algorithm for biomass monitoring
- Vecchia-Laplace approximations of generalized Gaussian processes for big non-Gaussian spatial data
- Stochastic variational inference for scalable non-stationary Gaussian process regression
- An efficient implementation for spatial-temporal Gaussian process regression and its applications
- A nonstationary space-time Gaussian process model for partially converged simulations
- A scalable approximate Bayesian inference for high-dimensional Gaussian processes
- Scaling relationships of Gaussian processes
- Gaussian process learning via Fisher scoring of Vecchia's approximation
- A level-crossing-based scaling dimensionality transform applied to stationary Gaussian processes
- Posterior inference for sparse hierarchical non-stationary models
- Fast adaptive Fourier integration for spectral densities of Gaussian processes
- Stochastic approximation of score functions for Gaussian processes
- Fast spatial Gaussian process maximum likelihood estimation via skeletonization factorizations
This page was built for publication: Scalable computations for nonstationary Gaussian processes
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6173565)
