Feature Space Adaptation for Effortless Gaussian Process Flows
Abstract
Outside the linear-Gaussian regime, conditional sampling from Gaussian processes (GPs) is challenging. Recent methods such as FlowGP (Moss et al., (2026)) can condition on arbitrary non-linear and non-Gaussian statements, but at considerable cost: an expensive iterative and high-dimensional diffusion that requires hand-specified kernel hyperparameters. In this paper, we alleviate two significant drawbacks of FlowGP by (1) introducing kernel approximations that enable scaling to high-resolution domains and (2) proposing a way to obtain the marginal likelihood by measuring the work needed to steer the diffusion towards conditioning statements. We enable, for the first time, hyperparameter optimisation within FlowGP and demonstrate our approach on probabilistic downscaling from areal summary statistics, PDE solution inference on irregular domains, and recovery of sea level anomaly fields from non-Gaussian satellite observations.
Related
- Temporal Fourier Likelihoods with Spatial Hilbert-Space Gaussian Process Approximations
- Higher-Order Gaussian Approximations on Path Space for the Simple Symmetric Exclusion Process
- A Gaussian Process State Space Formulation for Materials Constitutive Modeling: Application to Quasi-brittle Materials
- Testing Gaussian process with applications to super-resolution
- A Gaussian Process Regression-Enhanced Two-Sweep Parameterized Power Method for Computing Extreme Eigenpairs
BibTeX
@misc{cowperthwaite2026feature,
title = {{Feature Space Adaptation for Effortless Gaussian Process Flows}},
author = {Thomas Cowperthwaite and Louis Sharrock and Lachlan Astfalck and Henry Moss},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.11459},
eprint = {2610.11459},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.11459},
note = {CC BY-SA 4.0},
}
RIS
TY - UNPB
TI - Feature Space Adaptation for Effortless Gaussian Process Flows
AU - Thomas Cowperthwaite
AU - Louis Sharrock
AU - Lachlan Astfalck
AU - Henry Moss
PY - 2026
DA - 2026/10/08
PB - arXiv
DO - 10.48550/arxiv.2610.11459
UR - https://arxiv.org/abs/2610.11459
AB - Outside the linear-Gaussian regime, conditional sampling from Gaussian processes (GPs) is challenging. Recent methods such as FlowGP (Moss et al., (2026)) can condition on arbitrary non-linear and non-Gaussian statements, but at considerable cost: an expensive iterative and high-dimensional diffusion that requires hand-specified kernel hyperparameters. In this paper, we alleviate two significant drawbacks of FlowGP by (1) introducing kernel approximations that enable scaling to high-resolution domains and (2) proposing a way to obtain the marginal likelihood by measuring the work needed to steer the diffusion towards conditioning statements. We enable, for the first time, hyperparameter optimisation within FlowGP and demonstrate our approach on probabilistic downscaling from areal summary statistics, PDE solution inference on irregular domains, and recovery of sea level anomaly fields from non-Gaussian satellite observations.
ER -