Binary genetic algorithm using EDA-based problem decomposition for large-scale global optimization

In recent years many real-world optimization problems have had to deal with growing dimensionality. Such optimization problems, that are called large-scale global optimization (LSGO) problems, contain many hundreds or thousands of variables and are not separable. Moreover, many real-world problems are usually complex for detailed analysis, thus they are viewed as the black-box optimization problems. Thus, we can use the “blind” search techniques only. The most advanced techniques for LSGO are population-based stochastic search algorithms and are based on cooperative coevolution schemes using the problem decomposition via variables. These algorithms are mainly proposed for the real-valued search space and cannot be applied for problems with discrete or mixed variables. In this paper a novel technique is proposed, that uses a combination of a binary genetic algorithm (GA) and an estimation of distribution algorithm (EDA). The GA is used for solving the main optimization problem, and the EDA is used for collecting statistical data based on the past search experience to provide the problem decomposition by fixing perspective genes in chromosomes. The proposed EDA-based decomposition technique has the benefits of two general LSGO concepts: the random grouping methods and the dynamic learning methods. A standard implementation of the EDA-based decomposition GA and an implementation using the island model for parallel computing are discussed. The results of numerical experiments for benchmark problems from the IEEE CEC competition on the LSGO are presented. The experiments show that the approach demonstrates efficiency comparable to other advanced techniques. At the same time, the proposed approach can be applied for LSGO problems with arbitrary variables as it uses the binary representation of solutions.