A Hierarchical Hybrid Evolutionary Computation for Continuous Function Optimization

In this paper, we propose a hybrid master/slave approach to optimization problems on the basis of estimation of distribution algorithms (EDAs) and genetic algorithms (GAs). The master process estimates the probability distribution of the search space on the basis of the non-dependency model at each iteration and sends probability vectors to the slaves. The slaves use the vectors to generate a new initial population for their GA operations. We employ the simplest probability models and we compensate for the reduced accuracy problems by applying GAs to the solutions sampled using the simplest model. Moreover, our method can be incorporated with strategy research, and it easily can be parallelized. Lastly, we conduct experiments to verify the effectiveness of our method.