A Zoning Search Strategy for Differential Evolution Variants

摘要

Computational time and solution precision are two major concerns in the evolutionary computation (EC). To save the computational time, various high-performance computing techniques have been successfully employed to help meta-heuristic algorithms. Moreover, most of previous studies have focused on the performance improvement of meta-heuristic algorithms for improving the solution precision. However, the use of high- performance computing techniques and the improvement of algorithm performance do not mean that meta-heuristic algorithms can locate a promising region in some cases due to their search capability limitation. Therefore, converting improved computation power indirectly into algorithm search capability is a vital task. To achieve the above goal, a zoning search (ZS) strategy is proposed in the current work. In ZS, some variables of an optimization problem are selected randomly, and then they are divided into some segments, i.e., the entire search space of the optimization problem is divided into some subspaces. Finally, a selected meta-heuristic algorithm is used to find a satisfactory solution in each subspace. The effectiveness of ZS is demonstrated on IEEE CEC2014. Results suggest that ZS is a highly competitive approach to solve complex optimization problems.