摘要: 为了有效检测多目标优化进化算法的性能,从3个方面进行多目标优化测试问题的设计,即约束条件、最优解分布的均匀性、算法逼近Pareto最优前沿的难度,采用NSGA-Ⅱ算法对这些测试问题进行仿真实验,并将算法求得的最优解可视化。结果显示,测试问题能够有效检测算法在上述3方面的性能。
关键词:
多目标优化,
进化算法,
Pareto最优,
测试问题
Abstract: In order to test and evaluate the performance of Multi-Objective Evolutionary Algorithm(MOEA), multi-objective optimization test problems are suggested in this paper on the following perspectives: constrained condition, uniform representation of Pareto-optimal solutions and hindrance to reach the global Pareto-optimal front. NSGA-Ⅱ is used to make experiments on these test problems and the non-dominated fronts are visualized. Test results show that these problems can test the algorithm’s performance effectively in above three aspects.
Key words:
multi-objective optimization,
evolutionary algorithms,
Pareto-optimality,
test problems
中图分类号:
程 鹏;张自力. 多目标进化算法测试问题的设计与分析[J]. 计算机工程, 2009, 35(14): 238-240.
CHENG Peng; ZHANG Zi-li. Design and Analysis of Test Problems for Multi-Objective Evolutionary Algorithms[J]. Computer Engineering, 2009, 35(14): 238-240.