The performance of cohort genetic algorithms on royal road function and multi-modal functions

The hitchhiking is a well-known phenomenon for global optimization by genetic algorithms (GAs). This is a crucial factor which induces premature convergence in GAs. Cohort genetic algorithm (CGA) is a multipopulation algorithm proposed by Holland, which is used to explore search spaces for building blocks by Hyper-plane Defined Functions. In this paper, the convergence proof of CGA is presented. The parameter control of CGA is the key to improve performance of the algorithm. Moreover, maintaining population diversity is an important strategy to avoid premature convergence and escape from local optimum. So, different values of parameters are checked to measure the algorithm performance. In this paper, on the basis of a novel definition of population diversity, CGA is used to optimize the royal road function (RR) and multi-modal functions. The results show that the performance of CGA on multi-modal functions is better than on RR. Furthermore, an explanation is addressed for this phenomenon in the view of schema evolution.