Two Dimensional Evaluation Scheme in Genetic Algorithms
Sung Hoon Jung · Journal of Electrical Engineering and Information Science · 1999
This paper proposes a two dimensional evaluation scheme in genetic algorithms for improving their evolution ability. The main idea of this paper is to take into account not only the fitness of individuals, but also their improved fitness in the evaluation of genetic algorithms. This evaluation operation provides some individuals---whose fitness values are relatively small. but whose improved fitness values are considerably large---with more chances to take part in the evolution. This evaluation operation prevents genetic algorithms from falling into a premature convergence phenomenon especially in complex optimization problems with many local maxima. The performances of genetic algorithms with the original and proposed evaluation scheme were compared with five function optimization problems and one combination problem. It was found through extensive experiments that genetic algorithms with the proposed evaluation scheme outperform those with the original evaluation scheme. This evaluation scheme can be incorporated into other genetic algorithms to improve their performances.