Counter-niching for constructive population diversity

Maintaining a desired level of diversity in the evolutionary algorithm (EA) population is a requirement to ensure that progress of the EA search is unhindered by premature convergence to suboptimal solutions. Loss of diversity in the EA population pushes the search to a state where the genetic operators can no longer produce superior or even different offspring required to escape the local optimum. Besides diversity's contribution to avoid premature convergence, it is also useful to locate multiple optima where there is more than one solution available. This paper presents a counter-niching technique to introduce and maintain constructive diversity in the EA population. The proposed technique presented here uses informed genetic operations to reach promising, but un-explored or under-explored areas of the search space, while discouraging premature local convergence. Elitism is used at a different level aiming at convergence. The proposed technique's improved performance in terms solution accuracy and computation time is observed through simulation runs on a number of standard benchmark test functions with a genetic algorithm (GA) implementation.