A two-step selection scheme for constrained evolutionary optimization
Ming Chang, Kazuhiro Ohkura, K. Ueda, Masaharu Sugiyama · 2003
In biology, the fitness of an organism includes both its ability to survive and its ability to reproduce. However, most selection schemes used in evolutionary algorithms (EAs) have only embraced half part of this: the matter of concern is fertility in genetic algorithms (GA) in general, and viability in evolution strategies (ES) and evolutionary programming (EP) in particular. Although selection schemes that impose selection pressure on both viability and fertility exist as a minority, they have been mainly applied to single-objective optimization, in which viability and fertility are both evaluated according to the one and only objective function value. In this paper, we described a two-step selection scheme for constrained evolutionary optimization: viability selection and fertility selection procedures are executed sequentially during the life cycle of individuals, where viability and fertility are evaluated according to penalty function and objective respectively. The experimental results on thirteen benchmark problems show that the new selection scheme is quite comparable to the other state-of-the-art schemes.