A Comparison of the Effects of Dominance on Evolutionary Programming and Genetic Algorithms

Genetic Algorithms (GA) and Evolutionary Programming(EP) are both search techniques predicated upon simulating some of the processes outlined in evolutionary theories. Both techniques claim the capacity to find global minima (or alternately maxima) in large search spaces. Genetic Algorithms employ sexual reproduction among fit parents using genotypic operators such as crossover, bit mutation, and inversion. On the other hand, Evolutionary Programming espouses a form of asexual phenotypic reproduction whereby each candidate parent undergoes a zero-mean Gaussian mutation. Previous studies have shown that EP is capable of producing more precise solutions than GAs for a number of simple functions. This paper investigates another class of problem spaces for which GA outperforms EP, specifically problems where portions of a given solution have a greatly varying impact on the overall score. A simple representative problem is tested against typical EP and GA implementations with special attention paid to testing the effects of search parameters on both EP and GA performance. Finally, the relative performance and ease of use of the two paradigms is compared.