An adaptive simplex genetic algorithm
Linyu Yang, John Yen · 2000
In this paper, based on our previous research on simplex genetic algorithm, we put forward an adaptive approach for the self-adaptation of the percentage of simplex. According to the average fitness of those individuals generated by simplex operator in a certain generation, a set of rules is designed to adaptively adjust the simplex percentage within a certain range. This helps the algorithm always run at a relative optimized simplex percentage. We test our approach on both testbed problems and a real metabolic modeling problem. The results are satisfactory. 1 SIMPLEX-GA Simplex method is a local search technique that uses the current data set to determine the promising direction of search. We developed a simplex GA hybrid approach by introducing the simplex method as an additional operator in the genetic algorithm. (Yen et al. 1998). During the reproduction step of each iteration, the hybrid approach applies the simplex method to a top percentage of the population to produce new candidate solutions in the next generation. The rest of the new population are generated using the GA’s reproduction scheme (i.e., selection, crossover, and mutation). Before the run, we should specify the simplex percentage, which corresponds to the percentage of population to which the simplex is applied. For example, a 50 % simplex-GA applies the simplex to the top half population.