Barely Biased Learning for Gaussian Process Regression

David R. Burt, Artem Artemev, Mark van der Wilk

Keywords: Gaussian Process, Empirical Bayes, Regression

TL;DR: We propose an adaptive method for estimating the log marginal likelihood of conjugate Gaussian process regression with guarantees.

Abstract: Recent work in scalable approximate Gaussian process regression has discussed a bias-variance-computation trade-off when estimating the log marginal likelihood. We suggest a method that adaptively selects the amount of computation to use when estimating the log marginal likelihood so that the bias of the objective function is guaranteed to be small. While in principle a simple modification of existing approximations, our current implementation of the method is not computationally competitive with these existing approximations, limiting its applicability.

Category: Stuck paper: I hope to get ideas in this workshop that help me unstuck and improve this paper