BayeSQP: Bayesian Optimization through Sequential Quadratic Programming

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Brunzema, Paul, Trimpe, Sebastian
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911418130890752
author Brunzema, Paul
Trimpe, Sebastian
author_facet Brunzema, Paul
Trimpe, Sebastian
contents We introduce BayeSQP, a novel algorithm for general black-box optimization that merges the structure of sequential quadratic programming with concepts from Bayesian optimization. BayeSQP employs second-order Gaussian process surrogates for both the objective and constraints to jointly model the function values, gradients, and Hessian from only zero-order information. At each iteration, a local subproblem is constructed using the GP posterior estimates and solved to obtain a search direction. Crucially, the formulation of the subproblem explicitly incorporates uncertainty in both the function and derivative estimates, resulting in a tractable second-order cone program for high probability improvements under model uncertainty. A subsequent one-dimensional line search via constrained Thompson sampling selects the next evaluation point. Empirical results show thatBayeSQP outperforms state-of-the-art methods in specific high-dimensional settings. Our algorithm offers a principled and flexible framework that bridges classical optimization techniques with modern approaches to black-box optimization.
format Preprint
id arxiv_https___arxiv_org_abs_2602_03232
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle BayeSQP: Bayesian Optimization through Sequential Quadratic Programming
Brunzema, Paul
Trimpe, Sebastian
Machine Learning
We introduce BayeSQP, a novel algorithm for general black-box optimization that merges the structure of sequential quadratic programming with concepts from Bayesian optimization. BayeSQP employs second-order Gaussian process surrogates for both the objective and constraints to jointly model the function values, gradients, and Hessian from only zero-order information. At each iteration, a local subproblem is constructed using the GP posterior estimates and solved to obtain a search direction. Crucially, the formulation of the subproblem explicitly incorporates uncertainty in both the function and derivative estimates, resulting in a tractable second-order cone program for high probability improvements under model uncertainty. A subsequent one-dimensional line search via constrained Thompson sampling selects the next evaluation point. Empirical results show thatBayeSQP outperforms state-of-the-art methods in specific high-dimensional settings. Our algorithm offers a principled and flexible framework that bridges classical optimization techniques with modern approaches to black-box optimization.
title BayeSQP: Bayesian Optimization through Sequential Quadratic Programming
topic Machine Learning
url https://arxiv.org/abs/2602.03232