Sparse experimental design: an effective an efficient way discovering better genetic algorithm structures

The focus of this paper is the demonstration that sparse experimental design is a useful strategy for developing Genetic Algorithms. It is increasingly apparent from a number of reports and papers within a variety of different problem domains that the 'best' structure for a GA may be dependent upon the application. The GA structure is defined as both the types of operators and the parameters settings used during operation. The differences observed may be linked to the nature of the problem, the type of fitness function, or the depth or breadth of the problem under investigation. This paper demonstrates that advanced experimental design may be adopted to increase the understanding of the relationships between the GA structure and the problem domain, facilitating the selection of improved structures with a minimum of effort.