Scalability of the Bayesian optimization algorithm

Abstract

To solve a wide range of different problems, the research in black-box optimization faces several important challenges. One of the most important challenges is the design of methods capable of automatic discovery and exploitation of problem regularities to ensure efficient and reliable search for the optimum. This paper discusses the Bayesian optimization algorithm (BOA), which uses Bayesian networks to model promising solutions and sample new candidate solutions. Using Bayesian networks in combination with population-based genetic and evolutionary search allows BOA to discover and exploit regularities in the form of a problem decomposition. The paper analyzes the applicability of the methods for learning Bayesian networks in the context of genetic and evolutionary search and concludes that the combination of the two approaches yields robust, efficient, and accurate search.