Modified cGA for electromagnetic optimization

Compact Genetic Algorithm (cGA), which uses a probability vector (PV) to represent the population, has been proposed as an alternative of the simple Genetic algorithm (sGA), which greatly reduces the memory storage requiring during its performance. The cGA, however, just performed equivalently to sGA. In this paper, a modified version of compact Genetic Algorithm (M-cGA), outperforming the standard cGA, is presented. The idea is to use more than one probability vector and add a suitable learning scheme to improve the cGA's capability. Numerical results of the application of M-cGA on high-order problem, i.e. four-bit problem, and electromagnetic optimization, i.e. thinned array synthesis, will be presented and compared with the results obtained by its ancestors and GA as well.