Computer Engineering and Applications ›› 2007, Vol. 43 ›› Issue (11): 75-77.

• 学术探讨 • Previous Articles     Next Articles

An Effective Multi-objective Evolutionary Algorithm

Bing   

  • Received:2006-05-16 Revised:1900-01-01 Online:2007-04-11 Published:2007-04-11
  • Contact: Bing

一种高效的多目标演化算法

黄樟灿 焉炳艳 谢啸虎   

  1. 武汉理工大学理学院
  • 通讯作者: 焉炳艳

Abstract: This paper proposes a novel muiti-objective evolutionary algorithm, based on a novel crossover operation and improved crowding operation of NSGA-II ,in order to quicken further rate of convergence of solutions to Pareto optimal front and improve precision of solutions. The numeric experiments results indicate the new algorithm is very efficient for multi-objective test problems of high-dimension with Pareto optimal front of convex or non-convex or discontinuous and convex and multi-modal. The obtained non-dominated solutions have a good distribution property. But as to high-dimension functions with too many local Pareto optimal fronts, it traps in local Pareto optimal front easily.

Key words: multi-objective optimization problem, muiti-objective evolutionary algorithm, Pareto optimality

摘要: 为了提高非劣解向Pareto最优前沿收敛的速度及进一步提高解的精度,本文在设计了一种新的杂交算子并改进了NSGA-II的拥挤操作的基础上,提出了一种基于分级策略的多目标演化算法。数值实验表明,新算法能够非常高效的处理高维的最优前沿为凸的、非凸的和不连续的多峰的多目标测试函数,得到的非劣解具有很好的分布性质。但在处理高维的具有太多局部最优前沿的函数时极易陷入局部最优前沿。

关键词: 多目标优化问题, 多目标演化算法, Pareto最优