Maximizing Reliability with Bayesian Optimization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Buckingham, Jack M., Couckuyt, Ivo, Branke, Juergen
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911416649252864
author Buckingham, Jack M.
Couckuyt, Ivo
Branke, Juergen
author_facet Buckingham, Jack M.
Couckuyt, Ivo
Branke, Juergen
contents Bayesian optimization (BO) is a popular, sample-efficient technique for expensive, black-box optimization. One such problem arising in manufacturing is that of maximizing the reliability, or equivalently minimizing the probability of a failure, of a design which is subject to random perturbations - a problem that can involve extremely rare failures ($P_\mathrm{fail} = 10^{-6}-10^{-8}$). In this work, we propose two BO methods based on Thompson sampling and knowledge gradient, the latter approximating the one-step Bayes-optimal policy for minimizing the logarithm of the failure probability. Both methods incorporate importance sampling to target extremely small failure probabilities. Empirical results show the proposed methods outperform existing methods in both extreme and non-extreme regimes.
format Preprint
id arxiv_https___arxiv_org_abs_2602_02432
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Maximizing Reliability with Bayesian Optimization
Buckingham, Jack M.
Couckuyt, Ivo
Branke, Juergen
Machine Learning
Optimization and Control
Bayesian optimization (BO) is a popular, sample-efficient technique for expensive, black-box optimization. One such problem arising in manufacturing is that of maximizing the reliability, or equivalently minimizing the probability of a failure, of a design which is subject to random perturbations - a problem that can involve extremely rare failures ($P_\mathrm{fail} = 10^{-6}-10^{-8}$). In this work, we propose two BO methods based on Thompson sampling and knowledge gradient, the latter approximating the one-step Bayes-optimal policy for minimizing the logarithm of the failure probability. Both methods incorporate importance sampling to target extremely small failure probabilities. Empirical results show the proposed methods outperform existing methods in both extreme and non-extreme regimes.
title Maximizing Reliability with Bayesian Optimization
topic Machine Learning
Optimization and Control
url https://arxiv.org/abs/2602.02432