A general framework for Vecchia approximations of Gaussian processes

Abstract

Gaussian processes (GPs) are commonly used as models for functions, time series, and spatial fields, but they are computationally infeasible for large datasets. Focusing on the typical setting of modeling data as a GP plus an additive noise term, we propose a generalization of the Vecchia (1988) approach as a framework for GP approximations. We show that our general Vecchia approach contains many popular existing GP approximations as special cases, allowing for comparisons among the different methods within a unified framework. Representing the models by directed acyclic graphs, we determine the sparsity of the matrices necessary for inference, which leads to new insights regarding the computational properties. Based on these results, we propose a novel sparse general Vecchia approximation, which ensures computational feasibility for large spatial datasets but can lead to considerable improvements in approximation accuracy over Vecchia's original approach. We provide several theoretical results and conduct numerical comparisons. We conclude with guidelines for the use of Vecchia approximations in spatial statistics.

References in corpus (5)

• Fixed-domain asymptotic properties of tapered maximum likelihood estimators
• Covariance approximation for large multivariate spatial data sets with an application to multiple climate model errors
• Vecchia approximations of Gaussian-process predictions
• A class of multi-resolution approximations for large spatial datasets
• When does the screening effect hold?

Cited by in corpus (23)

• Scalable Bayesian modeling for smoothing disease risks in large spatial data sets
• A Deep Learning Synthetic Likelihood Approximation of a Non-stationary Spatial Model for Extreme Streamflow Forecasting
• Big problems in spatio-temporal disease mapping: methods and software
• Modeling Extremal Streamflow using Deep Learning Approximations and a Flexible Spatial Process
• High-dimensional order-free multivariate spatial disease mapping
• Neural Likelihood Surfaces for Spatial Processes with Computationally Intensive or Intractable Likelihoods
• Correlation-based sparse inverse Cholesky factorization for fast Gaussian-process inference
• Scalable marginalization of correlated latent variables with applications to learning particle interaction kernels
• A Scalable Gaussian Process for Large-Scale Periodic Data
• Bayesian nonparametric generative modeling of large multivariate non-Gaussian spatial fields
• Iterative Methods for Vecchia-Laplace Approximations for Latent Gaussian Process Models
• Monte Carlo inference for semiparametric Bayesian regression
• The inverse Kalman filter
• Monotonic warpings for additive and deep Gaussian processes
• Incorporating Subsampling into Bayesian Models for High-Dimensional Spatial Data
• Approximate Reference Prior for Gaussian Random Fields
• Linear-Cost Vecchia Approximation of Multivariate Normal Probabilities
• A Low-Rank Bayesian Approach for Geoadditive Modeling
• Compactly-supported nonstationary kernels for computing exact Gaussian processes on big data
• Gridding and Parameter Expansion for Scalable Latent Gaussian Models of Spatial Multivariate Data
• Inverse sampling intensity weighting for preferential sampling adjustment
• A Framework for the Bayesian Calibration of Complex and Data-Scarce Models in Applied Sciences
• Inside-out cross-covariance for spatial multivariate data