ABSTRACT
It is necessary to consider what makes problems easy or hard for genetic algorithms. Davidor [Dav91] has examined this question by applying Epistasis Analysis. In this paper, we first give a review of Epistasis Analysis. Then, we apply Epistasis Analysis to certain functions and construct five functions with differing degrees of difficulty for genetic algorithms, some GA-easy and some GA-hard. Finally, experimental results and conclusion are described. Unfortunately, Epistasis Analysis can not accurately estimate the GA-hardness of a function.
