Scalable Gaussian Process Inference with Finite-data Mean and Variance Guarantees
share
Gaussian processes (GPs) offer a flexible class of priors for nonparametric Bayesian regression, but popular GP posterior inference methods are typically prohibitively slow or lack desirable finite-data guarantees on quality. We develop an approach to scalable approximate GP regression with finite-data guarantees on the accuracy of pointwise posterior mean and variance estimates. Our main contribution is a novel objective for approximate inference in the nonparametric setting: the "preconditioned Fisher (pF) divergence." We show that unlike popular divergences such as Kullback--Leibler (used in variational inference), the pF divergence bounds the 2-Wasserstein distance, which in turn bounds the pointwise difference of the mean and variance functions. We demonstrate that, for sparse GP likelihood approximations, we can minimize the pF divergence efficiently. Our experiments show that optimizing the pF divergence has the same computational requirements as variational sparse GPs while providing comparable empirical performance---in addition to our novel finite-data quality guarantees.
READ FULL TEXT