Improving Differential Evolution Algorithm by Synergizing Different Improvement Mechanisms

摘要

Differential Evolution (DE) is a well-known Evolutionary Algorithm (EA) for solving global optimization problems. Practical experiences, however, show that DE is vulnerable to problems like slow and/ or premature convergence. In this article we propose a simple and modified DE framework, called MDE, which is a fusion of three recent modifications in DE: (1) Opposition-Based Learning (OBL); (2) tournament method for mutation; and (3) single population structure. These features have a specific role which helps in improving the performance of DE. While OBL helps in giving a good initial start to DE, the use of the tournament best base vector in the mutation phase helps in preserving the diversity. Finally the single population structure helps in faster convergence. Their synergized effect balances the exploitation and exploration capabilities of DE without compromising with the solution quality or the convergence rate. The proposed MDE is validated on a set of 25 standard benchmark problems, 7 nontraditional shifted benchmark functions proposed at the special session of CEC2008, and three engineering design problems. Numerical results and statistical analysis show that the proposed MDE is better than or at least comparable to the basic DE and several other state-of-the art DE variants.