When are Iterative Gaussian Processes Reliably Accurate?
Machine Learning
2022-01-03 v1 Machine Learning
Abstract
While recent work on conjugate gradient methods and Lanczos decompositions have achieved scalable Gaussian process inference with highly accurate point predictions, in several implementations these iterative methods appear to struggle with numerical instabilities in learning kernel hyperparameters, and poor test likelihoods. By investigating CG tolerance, preconditioner rank, and Lanczos decomposition rank, we provide a particularly simple prescription to correct these issues: we recommend that one should use a small CG tolerance () and a large root decomposition size (). Moreover, we show that L-BFGS-B is a compelling optimizer for Iterative GPs, achieving convergence with fewer gradient updates.
Cite
@article{arxiv.2112.15246,
title = {When are Iterative Gaussian Processes Reliably Accurate?},
author = {Wesley J. Maddox and Sanyam Kapoor and Andrew Gordon Wilson},
journal= {arXiv preprint arXiv:2112.15246},
year = {2022}
}
Comments
ICML 2021 OPTML Workshop