Comparing Scale Parameter Estimators for Gaussian Process Interpolation with the Brownian Motion Prior: Leave-One-Out Cross Validation and Maximum Likelihood

Abstract

Gaussian process (GP) regression is a Bayesian nonparametric method for regression and interpolation, offering a principled way of quantifying the uncertainties of predicted function values. For the quantified uncertainties to be well-calibrated, however, the kernel of the GP prior has to be carefully selected. In this paper, we theoretically compare two methods for choosing the kernel in GP regression: cross-validation and maximum likelihood estimation. Focusing on the scale-parameter estimation of a Brownian motion kernel in the noiseless setting, we prove that cross-validation can yield asymptotically well-calibrated credible intervals for a broader class of ground-truth functions than maximum likelihood estimation, suggesting an advantage of the former over the latter. Finally, motivated by the findings, we propose interior cross validation, a procedure that adapts to an even broader class of ground-truth functions.

References in corpus (7)

• Reproducing kernel Hilbert spaces of Gaussian priors
• Fixed-domain asymptotic properties of tapered maximum likelihood estimators
• Frequentist coverage of adaptive nonparametric Bayesian credible sets
• Bayesian Probabilistic Numerical Methods
• Fixed-domain asymptotics for a subclass of Matérn-type Gaussian random fields
• A novel class of stabilized greedy kernel approximation algorithms: Convergence, stability & uniform point distribution
• Asymptotic properties of the maximum likelihood and cross validation estimators for transformed Gaussian processes