A novel multi-fidelity sequential optimization method based on multi-level Gaussian process

doi:10.1016/j.engappai.2025.112678

A novel multi-fidelity sequential optimization method based on multi-level Gaussian process

2025-10-16

0

PRE

AI

DOI：10.1016/j.engappai.2025.112678

Original

Original request for help

Save

Abstract

En 中文

• Proposes Multi-level Expected Improvement criterion (LEI) for joint sampling and fidelity optimization. • Scales to >2 fidelity levels, cutting 40 % high-fidelity samples in 5-level tasks. • Establishes MLGP convergence and LEI robustness; theory confirms low-fidelity data enhances high-fidelity accuracy. • In HEG test, LEI finds optimum with only 16 high-fidelity runs, slashing cost by 60 %.

A novel multi-fidelity sequential optimization method based on multi-level Gaussian process

Abstract

Journal

Organization