Optimum population size and mutation rate for a simple real genetic algorithm that optimizes array factors

There has been an explosion of papers describing applications of a genetic algorithm (GA) to electromagnetics problems. Most of the work has followed traditional GA philosophy when choosing the population size and mutation rate of the genetic algorithm. This paper reports the results of experiments to determine the optimum population size and mutation rate for a simple real genetic algorithm. The choice of population size and mutation rate can cause the run time of the GA to vary by several orders of magnitude. The results of this investigation show that a small population size and relatively large mutation rate is far superior to the large population sizes and low mutation rates that is used by most of the papers presented in the electromagnetics community and by the GA community at large. The results of the numerical experiments presented in this paper suggest that the best mutation rate for GAs lies between 5 and 20% while the population size should be less than 16.