SABLE: Scalable Affine Bayesian Latent-space Exploration
Abstract
Complex objects such as engineering designs and proteins are often difficult to optimise because their design spaces exhibit high-dimensional structure and objective evaluations are expensive. Generative models provide continuous latent representations, but their latent spaces can have thousands to tens of thousands of dimensions, creating a regime where the latent dimension D greatly exceeds the evaluation budget. We introduce SABLE, a Bayesian optimisation method for high-dimensional latent spaces with an isotropic distribution and limited budget. SABLE uses a Bayesian affine surrogate for prediction and acquisition, treating higher-order structure as surrogate discrepancy. After N observations, the slope update has dimension at most N−1, yielding a decomposition of each query into an in-span component optimised by expected improvement (EI) in at most N dimensions and an off-span component governed by a scalar norm. We show that the off-span norm implicitly selected by affine EI can lead to underexploration, and SABLE instead prescribes it using a finite-horizon schedule, without restricting search to a learned subspace. A recompression scheme reduces information to 𝒪(M) pseudo-observations, bounding refitting cost by 𝒪(M2D+M3) while keeping parameter inference, prediction, and acquisition decisions to tens of milliseconds. Across synthetic benchmarks and latent-space optimisation tasks, SABLE achieves high sample efficiency and reduces decision time by several orders of magnitude at large budgets.
Related
- Learning Directional Randers Geometry for Latent-Space Exploration
- Sharp Approximation Rates for Neural Networks with Affine Latent Parameterizations
- CABLE: Convex Affine Blends of Latent Experts for parametric reduced-order modeling
- Bayesian Sample Size Determination: Sampling Distribution Estimation or Exploration?
- Recursive moment updates for scalable Bayesian nonparametric regression
BibTeX
@misc{bodin2026sable,
title = {{SABLE: Scalable Affine Bayesian Latent-space Exploration}},
author = {Erik Bodin and Lisa Maria Kreusser and Donney Fan and Henry Moss and Geoff Pleiss and Nicolas Durrande},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.31495},
eprint = {2610.31495},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.31495},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - SABLE: Scalable Affine Bayesian Latent-space Exploration
AU - Erik Bodin
AU - Lisa Maria Kreusser
AU - Donney Fan
AU - Henry Moss
AU - Geoff Pleiss
AU - Nicolas Durrande
PY - 2026
DA - 2026/10/23
PB - arXiv
DO - 10.48550/arxiv.2610.31495
UR - https://arxiv.org/abs/2610.31495
AB - Complex objects such as engineering designs and proteins are often difficult to optimise because their design spaces exhibit high-dimensional structure and objective evaluations are expensive. Generative models provide continuous latent representations, but their latent spaces can have thousands to tens of thousands of dimensions, creating a regime where the latent dimension D greatly exceeds the evaluation budget. We introduce SABLE, a Bayesian optimisation method for high-dimensional latent spaces with an isotropic distribution and limited budget. SABLE uses a Bayesian affine surrogate for prediction and acquisition, treating higher-order structure as surrogate discrepancy. After N observations, the slope update has dimension at most N−1, yielding a decomposition of each query into an in-span component optimised by expected improvement (EI) in at most N dimensions and an off-span component governed by a scalar norm. We show that the off-span norm implicitly selected by affine EI can lead to underexploration, and SABLE instead prescribes it using a finite-horizon schedule, without restricting search to a learned subspace. A recompression scheme reduces information to 𝒪(M) pseudo-observations, bounding refitting cost by 𝒪(M^2D+M^3) while keeping parameter inference, prediction, and acquisition decisions to tens of milliseconds. Across synthetic benchmarks and latent-space optimisation tasks, SABLE achieves high sample efficiency and reduces decision time by several orders of magnitude at large budgets.
ER -