Understanding Hidden Spatial Preferences in Bayesian Optimization
Abstract
Bayesian optimization uses posterior uncertainty to decide where to explore, but this uncertainty is not spatially neutral. On bounded domains, Gaussian-process posterior variance is typically larger near faces, edges, and corners, giving variance-based acquisition functions a systematic boundary bias. This spatial preference is usually inherited from the surrogate rather than chosen as part of the exploration strategy. We separate these roles by keeping the predictive model fixed and modifying the spatial preference of the exploration term. We control the preference with spatial weight fields and amortize their construction with neural surrogates. Controlled experiments show that the Gaussian process' default boundary bias changes which parts of the domain are explored and which optima are reached. We compare three alternative exploration principles and find, across standard benchmark functions and stochastic robot-pushing tasks, that they can often lower regret, particularly early in the optimization. The spatial structure of posterior uncertainty is therefore not a neutral default: it shapes the exploration strategy and can materially affect optimization performance.
Related
- Empirical Bayes in Bayesian learning: understanding a common practice
- Impact analysis of institutional framework and policy preferences with Bayesian structural equation
- Handling Customer Preferences in Last-Mile Location-Routing: A Bilevel Optimization Approach
- Gated Spatial Hidden Markov Models for Classification in Histological Samples, With Application to Muscle Fiber‐Typing
- A Bayesian Hidden Markov Model with Singh--Maddala Emission Distributions for Heavy-Tailed Regime-Switching Time Series
BibTeX
@misc{bankestad2026understanding,
title = {{Understanding Hidden Spatial Preferences in Bayesian Optimization}},
author = {Maria Bånkestad and Jens Sjölund},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.14696},
eprint = {2610.14696},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.14696},
}
RIS
TY - UNPB
TI - Understanding Hidden Spatial Preferences in Bayesian Optimization
AU - Maria Bånkestad
AU - Jens Sjölund
PY - 2026
DA - 2026/10/10
PB - arXiv
DO - 10.48550/arxiv.2610.14696
UR - https://arxiv.org/abs/2610.14696
AB - Bayesian optimization uses posterior uncertainty to decide where to explore, but this uncertainty is not spatially neutral. On bounded domains, Gaussian-process posterior variance is typically larger near faces, edges, and corners, giving variance-based acquisition functions a systematic boundary bias. This spatial preference is usually inherited from the surrogate rather than chosen as part of the exploration strategy. We separate these roles by keeping the predictive model fixed and modifying the spatial preference of the exploration term. We control the preference with spatial weight fields and amortize their construction with neural surrogates. Controlled experiments show that the Gaussian process' default boundary bias changes which parts of the domain are explored and which optima are reached. We compare three alternative exploration principles and find, across standard benchmark functions and stochastic robot-pushing tasks, that they can often lower regret, particularly early in the optimization. The spatial structure of posterior uncertainty is therefore not a neutral default: it shapes the exploration strategy and can materially affect optimization performance.
ER -