A hybrid estimation of distribution algorithm with differential evolution for global optimization

In evolutionary algorithms, it is difficult to balance the exploration and exploitation. Usually, global search is utilized to find promising solutions, and local search is beneficial to the convergence of the solutions in the population. Combing different search strategies is a promising way to take advantages of different methods. Following the idea of DE/EDA, this paper proposes another way to combine estimation of distribution algorithm and differential evolution for global optimization. The basic idea is to choose either differential evolution or estimation of distribution algorithm for generating new trial solutions. To improve the algorithm performance, a local search strategy is used as well. The new approach, named as EDA/DE-EIG, is systematically compared with two state-of-art algorithms, and the experimental results show the advantages of our method.