A peak shape identification genetic algorithm with a radial basis function

We propose a peak shape identification GA (psi-GA), that searches the best peak of a fitness function and identifies its shape, that is, how sharp or blunt the peak is. We introduce a parameter of radial basis functions in a chromosome of a genetic algorithm to represent the shape of a peak. The modified fitness function is defined as an additional form of the original fitness function and the similarity between the shape of a peak and that of the radial basis function defined by a chromosome. The calculation of similarity allows to implement a local search easily as a by-product. The psi-GA also has the availability to search peaks of a fitness function selectively according to their shapes. The empirical results show that the psi-GA has both sufficient performance to search the best peak of the fitness function and to identify the shape of the peak.