A novel genetic algorithm based on multi-species

This paper presents a novel genetic algorithms based on multi-species with a new mutation operator. The method is designed for optimizing complex multimodal functions in which the standard genetic algorithms (SGA) always gets struck into a local optimum. The paper introduces a new structure based on main species and additional species to avoid the premature convergence of SGA. In this structure, the main species use a new mutation operator to keep the population diversity in the entire search space and acquire the fast increasing of better models, and the additional species are designed to get local optima in the specified regions. The storing of the research history and the communication between the main and additional species help to decrease the research space for acquiring a global optimum and several local optima. The experiments performed to optimize several complex multimodal functions have acquired good results.