Marginalising over Stationary Kernels with Bayesian Quadrature

Abstract

Marginalising over families of Gaussian Process kernels produces flexible model classes with well-calibrated uncertainty estimates. Existing approaches require likelihood evaluations of many kernels, rendering them prohibitively expensive for larger datasets. We propose a Bayesian Quadrature scheme to make this marginalisation more efficient and thereby more practical. Through use of the maximum mean discrepancies between distributions, we define a kernel over kernels that captures invariances between Spectral Mixture (SM) Kernels. Kernel samples are selected by generalising an information-theoretic acquisition function for warped Bayesian Quadrature. We show that our framework achieves more accurate predictions with better calibrated uncertainty than state-of-the-art baselines, especially when given limited (wall-clock) time budgets.

References in corpus (5)

• Kernel Methods on Riemannian Manifolds with Gaussian RBF Kernels
• Sampling for Inference in Probabilistic Models with Fast Bayesian Quadrature
• Improving the Gaussian Process Sparse Spectrum Approximation by Representing Uncertainty in Frequency Inputs
• Function-Space Distributions over Kernels
• Automated Model Selection with Bayesian Quadrature