Brain Storm Optimization Algorithm with Estimation of Distribution

Abstract

Brain Storm Optimization (BSO) algorithm is a new intelligence optimization algorithm, which is effective to solve the multi-modal, high-dimensional and large-scale optimization problems. However, when the BSO algorithm deals with the complex problems, there are still some disadvantages, such as the slow speed of the search algorithm for the late, premature convergence and easy to fall into local optimal solutions and so on. In order to solve these problems, a BSO algorithm with Estimation of Distribution (EDBSO) is proposed. Similarity as the DMBSO, the EDBSO algorithm is divided the discussion process into two parts, including intra-group discussion and inter-group discussion. The Estimation of Distribution algorithm in continuous domains, that is based on the variables subject to Gaussian distribution, is used to improve the process of inter-group discussion of DMBSO algorithm. In this paper, five benchmark functions are used to evaluate the search performance of EDBSO algorithm. In order to verify the convergence and accuracy of the EDBSO algorithm, the EDBSO algorithm is compared with 4 improved algorithm in different dimensions. The simulation results show that the EDBSO algorithm can effectively avoid to falling into the local optimum and prevent premature convergence of this algorithm, and it can find better optimal solutions stably. With the increase of the problem dimensions, the EDBSO algorithm which has better robustness is suitable for solving complex optimization problems.