Optimization via simulation using Gaussian process-based search

TL;DR: This paper proposes a new random search algorithm, called the Gaussian Process-based Search (GPS), which derives a sampling distribution from a fast fitted Gaussian process in each iteration of the algorithm, and shows that the sampling distribution has the desired properties and it can automatically balance the exploitation and exploration tradeoff.

Abstract: Random search algorithms are often used to solve optimization-via-simulation (OvS) problems. The most critical component of a random search algorithm is the sampling distribution that is used to guide the allocation of the search effort. A good sampling distribution can balance the tradeoff between the effort used in searching around the current best solution (which is called exploitation) and the effort used in searching largely unknown regions (which is called exploration). However, most of the random search algorithms for OvS problems have difficulties in balancing this tradeoff in a seamless way. In this paper we propose a new random search algorithm, called the Gaussian Process-based Search (GPS) algorithm, which derives a sampling distribution from a fast fitted Gaussian process in each iteration of the algorithm. We show that the sampling distribution has the desired properties and it can automatically balance the exploitation and exploration tradeoff.