Iterative Methods for Vecchia-Laplace Approximations for Latent Gaussian Process Models

Abstract

Latent Gaussian process (GP) models are flexible probabilistic non-parametric function models. Vecchia approximations are accurate approximations for GPs to overcome computational bottlenecks for large data, and the Laplace approximation is a fast method with asymptotic convergence guarantees to approximate marginal likelihoods and posterior predictive distributions for non-Gaussian likelihoods. Unfortunately, the computational complexity of combined Vecchia-Laplace approximations grows faster than linearly in the sample size when used in combination with direct solver methods such as the Cholesky decomposition. Computations with Vecchia-Laplace approximations can thus become prohibitively slow precisely when the approximations are usually the most accurate, i.e., on large data sets. In this article, we present iterative methods to overcome this drawback. Among other things, we introduce and analyze several preconditioners, derive new convergence results, and propose novel methods for accurately approximating predictive variances. We analyze our proposed methods theoretically and in experiments with simulated and real-world data. In particular, we obtain a speed-up of an order of magnitude compared to Cholesky-based calculations and a threefold increase in prediction accuracy in terms of the continuous ranked probability score compared to a state-of-the-art method on a large satellite data set. All methods are implemented in a free C++ software library with high-level Python and R packages.

References in corpus (9)

• GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration
• A general framework for Vecchia approximations of Gaussian processes
• A class of multi-resolution approximations for large spatial datasets
• Vecchia approximations of Gaussian-process predictions
• Vecchia-Laplace approximations of generalized Gaussian processes for big non-Gaussian spatial data
• Scalable Log Determinants for Gaussian Process Kernel Learning
• Constant-Time Predictive Distributions for Gaussian Processes
• Preconditioning for Scalable Gaussian Process Hyperparameter Optimization
• When are Iterative Gaussian Processes Reliably Accurate?