Research On Coevolutionary Algorithms Using Mixed Strategies

Coevolutionary algorithm is a hot research in computational intelligence, whichaims at improving conventional evolutionary algorithms. Coevolution refers to evolv-ing two or more populations simultaneously. Through interactions among populations,e.g., competition and cooperation among populations and individuals, they may en-hance their performance and adaptation in more complex dynamic environment, andobtain a better survival and reproduction. Inspired by game theory, a new approachof designing coevolutionary algorithms with mixed strategies is proposed in the dis-sertation, that is, to design artificial game system among individuals and populations.This kind of new algorithms provides a new way to overcome the shortcoming of slowconvergence in conventional evolutionary algorithms. In the dissertation, game theoryinspired coevolutionary algorithms constructed and implemented; several experimentshave been finished. Evolutionary algorithms have been applied in the field of clustering,a new evolutionary clustering algorithm is proposed. The details are listed below:1. Inspired by evolutionary game theory, this dissertation presents a coevolution-ary programming with mixed strategies (MSCEP) that employs the Gaussian, Cauchy,L′evy, and single-point mutation operators. The novel algorithm is tested on a set of22 benchmark problems. The results show that the mixed strategy performs equallywell or better than the best of the four pure strategies does, for all of the benchmarkproblems.2. A mixed strategies evolutionary programming to solve constrained optimizationproblems is presented. The approach does not require the use of a penalty function.Instead, it uses a diversity conservation mechanism based on allowing infeasible solu-tions to remain in the population. A mixed mutation strategy and feasibility-basedcomparison mechanism is used to guide the process fast toward the feasible region...