A REAL-CODED CELLULAR GENETIC ALGORITHM INSPIRED BY PREDATOR-PREY INTERACTIONS

摘要

This chapter presents a real-coded cellular GA model using a new selection method inspired by predator-prey interactions. The model relies on the dynamics generated by spatial predator-prey interactions to maintain an appropriate selection pressure and diversity in the prey population. In this model, prey, which represent potential solutions, move around on a two-dimensional lattice and breed with other prey individuals. The selection pressure is exerted by predators, which also roam around to keep the prey in check by removing the weakest prey in their vicinity. This kind of selection pressure efficiently drives the prey population to greater fitness over successive generations. Our preliminary study has shown that the predator-prey interaction dynamics play an important role in maintaining an appropriate selection pressure in the prey population, thereby helping to generate suitably fit prey solutions. Our experimental results are comparable or better in performance than those of a standard serial and distributed real-coded GA.