Genetic algorithm with geographic speciation

Genetic algorithm (GA) is a bionic algorithm which is widely used for optimization problems. It was initially coined by professor Holland in University of Michigan who took advantage of some phenomenon in natural evolution, such as crossover, mutation, selection and inheritance. However, there is a longstanding problem of genetic algorithm. During the process of optimization, prematurity may occur. It means that local optimum, instead of the global one, is found. But when avoiding prematurity, another problem (i.e. converging slowly) will be triggered. This paper is aiming to propose a modified genetic algorithm simulating the geographic speciation (GS) which has an important role in species´ evolution. GS redirect species´ evolution by dividing a species into different parts and letting each one evolve in their particular area into a new species. In the meanwhile, GS keep accessing those species. This process is to evaluate whether a specific species is likely to evolve into a superior one. Some new species will be further divided when some conditions are satisfied. Otherwise, these species will no longer be studied. The underlying idea of the genetic algorithm with GS is dropping the domains that global optimal is not exist and investigating the domains that a believed global optimal, may be local, exist at a more precise level.