An Analysis of Diploidy and Dominance in Genetic Algorithms

The use of diploidy and dominance in genetic algorithms (GAs) has long been used to improve performance in time-varying optimization problems. Diploidy increases diversity in GAs by allowing recessive genes to survive in a population and become active at some later time when changes in the environment make them more desirable. This paper suggests an intuitive way to implement diploidy and presents some mathematical analyses of fitness proportional selection to justify its use in time-varying problems. An extension of the classical schema theorem for diploid GAs is presented. The mathematical analyses are geared towards the One Max problem, and assume a GA with selection and mutation only (no crossover). The analyses confirm that diploidy increases diversity, and provide some quantitative results for diversity increase as a function of the GA population characteristics.