Abstract:Large-scale Gaussian process models are becoming increasingly important and widely used in many areas, such as, computer experiments, stochastic optimization via simulation, and machine learning using Gaussian processes. The standard methods, such as maximum likelihood estimation (MLE) for parameter estimation and the best linear unbiased predictor (BLUP) for prediction, are generally the primary choices in many applications. In spite of their merits, those methods are not feasible due to intractable computation when the sample size is huge. A novel method for the purposes of parameter estimation and prediction is proposed to solve the computational problems of large-scale Gaussian process based models, by separating the original dataset into tractable subsets. This method consistently combines parameter estimation and prediction by making full use of the dependence among conditional densities: a statistically efficient composite likelihood based on joint distributions of some well selected conditional densities is developed to estimate parameters and then "composite inference" is coined to make prediction for an unknown input point, based on its distributions conditional on each block subset. The proposed method transforms the intractable BLUP into a tractable convex optimization problem. It is also shown that the prediction given by the proposed method, called the best linear unbiased block predictor, has a minimum variance for a given separation of the dataset. Keywords: Large scale, Parallel computing, Composite likelihood, Spatial process

Universal inference with composite likelihoods

When Composite Likelihood Meets Stochastic Approximation

Parsimonious and Efficient Likelihood Composition by Gibbs Sampling

Composite likelihood inference for the Poisson log-normal model

Bayesian Inference from Composite Likelihoods, with an Application to Spatial Extremes

A composite likelihood ratio approach to the analysis of correlated binary data in genetic association studies

On divergences tests for composite hypotheses under composite likelihood

Composite Likelihood Estimation for Restricted Boltzmann machines

Bayesian inference for Gibbs random fields using composite likelihoods

Near Universal Consistency of the Maximum Pseudolikelihood Estimator for Discrete Models

A Law of Likelihood for Composite Hypotheses

Composite Inference for Gaussian Processes

Robustness properties of marginal composite likelihood estimators

Statistical inference for the nonparametric and semiparametric hidden Markov model via the composite likelihood approach

Better Together: pooling information in likelihood-free inference

Multiple Comparisons Using Composite Likelihood in Clustered Data

Empirical Likelihood for Composite Quantile Regression Modeling

On universal inference in Gaussian mixture models

High-dimensional empirical likelihood inference

Composite likelihood and maximum likelihood methods for joint latent class modeling of disease prevalence and high-dimensional semicontinuous biomarker data

Generalised likelihood profiles for models with intractable likelihoods