Model Guided Sampling Optimization for Low-dimensional Problems

TL;DR

This paper introduces MGSO, a robust optimization method that uses sampling of the probability of improvement to efficiently optimize expensive black-box functions, outperforming traditional methods on low-dimensional problems.

Contribution

MGSO offers a novel sampling-based approach that enhances robustness and convergence speed compared to Gaussian-process-based methods like EGO.

Findings

MGSO reaches near-optimal solutions faster on low-dimensional problems.

Sampling the probability of improvement helps avoid local minima.

MGSO outperforms standard algorithms in efficiency and robustness.

Abstract

Optimization of very expensive black-box functions requires utilization of maximum information gathered by the process of optimization. Model Guided Sampling Optimization (MGSO) forms a more robust alternative to Jones' Gaussian-process-based EGO algorithm. Instead of EGO's maximizing expected improvement, the MGSO uses sampling the probability of improvement which is shown to be helpful against trapping in local minima. Further, the MGSO can reach close-to-optimum solutions faster than standard optimization algorithms on low dimensional or smooth problems.