Robust and Conjugate Gaussian Process Regression

2023/11/01 by Matias Altamirano, François‐Xavier Briol, François-Xavier Briol +4 · 1 voice · 5 citations
Biochemistry, Genetics and Molecular Biology · Chemistry · Computer Science · Mathematics · #FOS: Computer and information sciences #Gaussian Processes and Bayesian Inference #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Spectroscopy Techniques in Biomedical and Chemical Research #Spectroscopy and Chemometric Analyses #cs.LG #stat.ML

paper · pdf · doi:10.48550/arxiv.2311.00463

openalex publication_date 2023/11/01 · arxiv published 2023/11/01 · arxiv updated 2024/06/03 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

To enable closed form conditioning, a common assumption in Gaussian process (GP) regression is independent and identically distributed Gaussian observation noise. This strong and simplistic assumption is often violated in practice, which leads to unreliable inferences and uncertainty quantification. Unfortunately, existing methods for robustifying GPs break closed-form conditioning, which makes them less attractive to practitioners and significantly more computationally expensive. In this paper, we demonstrate how to perform provably robust and conjugate Gaussian process (RCGP) regression at virtually no additional cost using generalised Bayesian inference. RCGP is particularly versatile as it enables exact conjugate closed form updates in all settings where standard GPs admit them. To demonstrate its strong empirical performance, we deploy RCGP for problems ranging from Bayesian optimisation to sparse variational Gaussian processes.

Cited by

• Robust Bayesian Optimisation with Unbounded Corruptions
• Multi-Output Robust and Conjugate Gaussian Processes
• Distributionally Robust Causal Abstractions
• Decision Making under Model Misspecification: DRO with Robust Bayesian Ambiguity Sets
• Robust and Computation-Aware Gaussian Processes
• Joint Model and Data Sparsification via the Marginal Likelihood

Discussions

• The paper builds on the robust and scalable Gaussian process (RCGP) algorithm from our ICML 2024 paper (arxiv.org/abs/2311.00463). It shows that it can make use of common computational tricks in spati [bsky, 2 points, 1 comments]

Related

• Multi-Output Robust and Conjugate Gaussian Processes
• Robust and Computation-Aware Gaussian Processes
• Gaussian Process Regression and Classification under Mathematical Constraints with Learning Guarantees
• Nonnegativity-Enforced Gaussian Process Regression
• Robust and Conjugate Spatio-Temporal Gaussian Processes