An Experimental Design Perspective on Genetic Algorithms

In this paper we examine the relationship between genetic algorithms (GAs) and traditional methods of experimental design. This was motivated by an investigation into the problem caused by epistasis in the implementation and application of GAs to optimization problems: one which has long been acknowledged to have an important innuence on GA performance. Davidor 1, 2] has attempted an investigation of the important question of determining the degree of epistasis of a given problem. In this paper, we shall rst summarise his methodology, and then provide a critique from the perspective of experimental design. We proceed to show how this viewpoint enables us to gain further insights into the determination of epistatic eeects, and into the value of diierent forms of encoding a problem for a GA solution. We also demonstrate the equivalence of this approach to the Walsh transform analysis popularized by Goldberg 3, 4], and its extension to the idea of partition coeecients 5]. We then show how the experimental design perspective helps to throw further light on the nature of deception.