摘要
考虑到在基于分解的多目标进化算法(MOEA/D)中,邻域大小与变异算子类型对算法进化过程中的探索模式有不同的影响,提出优化的MOEA/D算法。4种不同大小的邻域范围和4个特性不同的变异策略两两组合构成候选池,利用负反馈原则,在进化过程中以较高概率从候选池中选择表现更优的组合。实验结果表明,该算法鲁棒性较强,在保证收敛性的同时具有较好的多样性。
Considering that the range of neighborhood sizes and the type of mutation operators have huge effect on the exploration mode in the algorithm evolution process in Multi-objective Evolutionary Algorithm based on Decomposition ( MOEA/D), this paper proposes an optimized MOEA/D algorithm. Four different neighborhood sizes and four mutation strategies with different features are combined in pairs as a part of candidate pool. In the evolutionary process, the combination with better performance is selected from the candidate pool with higher possibility according to the principle of negative feedback. Experimental results indicate that the proposed algorithm has strong robustness, and good diversity while ensuring convergence.
出处
《计算机工程》
CAS
CSCD
北大核心
2017年第3期232-240,共9页
Computer Engineering
基金
广东省对外科技合作基金(2013B051000060)
广东省教育部产学研结合基金重点项目(2011A090200085)
深圳市科技创新委员会基金(ZYC201105180515A)
关键词
邻域范围
变异算子类型
候选池
基于分解的多目标进化算法
多目标优化
neighborhood range
mutation operator type
candidate pool
Multi-objective Evolutionary Algorithm based on Decomposition (MOEA/D)
multi-objective optimization