Gaussian process learning via Fisher scoring of Vecchia's approximation
From MaRDI portal
Publication:125192
DOI10.48550/ARXIV.1905.08374zbMATH Open1475.62033arXiv1905.08374OpenAlexW3134488295MaRDI QIDQ125192FDOQ125192
Authors: Joseph Guinness 
Publication date: 20 May 2019
Published in: Statistics and Computing (Search for Journal in Brave)
Abstract: We derive a single pass algorithm for computing the gradient and Fisher information of Vecchia's Gaussian process loglikelihood approximation, which provides a computationally efficient means for applying the Fisher scoring algorithm for maximizing the loglikelihood. The advantages of the optimization techniques are demonstrated in numerical examples and in an application to Argo ocean temperature data. The new methods are more accurate and much faster than an optimization method that uses only function evaluations, especially when the covariance function has many parameters. This allows practitioners to fit nonstationary models to large spatial and spatial-temporal datasets.
Full work available at URL: https://arxiv.org/abs/1905.08374
Recommendations
- Scalable computations for nonstationary Gaussian processes
- Fast spatial Gaussian process maximum likelihood estimation via skeletonization factorizations
- Stochastic approximation of score functions for Gaussian processes
- Correlation-based sparse inverse Cholesky factorization for fast Gaussian-process inference
- Covariance tapering for likelihood-based estimation in large spatial data sets
Cites Work
- Approximating Likelihoods for Large Spatial Data Sets
- Inconsistent Estimation and Asymptotically Equal Interpolations in Model-Based Geostatistics
- A Simplex Method for Function Minimization
- A general framework for Vecchia approximations of Gaussian processes
- Scalable Gaussian Process Computations Using Hierarchical Matrices
Cited In (18)
- Nearest neighbors weighted composite likelihood based on pairs for (non-)Gaussian massive spatial data with an application to Tukey-\(hh\) random fields estimation
- Large-scale environmental data science with ExaGeoStatR
- Distributed Bayesian inference in massive spatial data
- Scalable semiparametric spatio-temporal regression for large data analysis
- A functional-data approach to the Argo data
- Scalable computations for nonstationary Gaussian processes
- GpGp
- Efficient Stochastic Generators with Spherical Harmonic Transformation for High-Resolution Global Climate Simulations from CESM2-LENS2
- Scaled Vecchia Approximation for Fast Computer-Model Emulation
- Vecchia-approximated Deep Gaussian Processes for Computer Experiments
- The third competition on spatial statistics for large datasets
- Fast adaptive Fourier integration for spectral densities of Gaussian processes
- Implementation and analysis of GPU algorithms for Vecchia approximation
- Stochastic approximation of score functions for Gaussian processes
- Fitting Matérn smoothness parameters using automatic differentiation
- Discussion on competition on spatial statistics for large datasets
- Correlation-based sparse inverse Cholesky factorization for fast Gaussian-process inference
- Hierarchical Bayesian modeling of ocean heat content and its uncertainty
This page was built for publication: Gaussian process learning via Fisher scoring of Vecchia's approximation
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q125192)
