A diversity-control-oriented genetic algorithm (DCGA): performance in function optimization

In genetic algorithms, in order to attain the global optimum without getting stuck at a local optimum, an appropriate diversity of structures in the population needs to be maintained. I have proposed a new genetic algorithm called DCGA (diversity control-oriented genetic algorithm) to attain this goal. In DCGA, the structures of the population in the next generation are selected from the merged population of parents and their offspring on the basis of a particular selection probability to maintain the diversity of the structures. The major feature is that the distance between a structure and the best performance structure is used as the primary selection criterion and it is applied on the basis of a probabilistic function that produces a larger selection probability for a structure with a larger distance. The performance of DCGA in function optimization is examined by experiments on benchmark problems. Within the range of my experiments, DCGA showed superior performance and it seems to be a promising competitor of the previously proposed algorithm