电子学报
電子學報
전자학보
ACTA ELECTRONICA SINICA
2015年
8期
1488-1498
,共11页
多目标优化%捕获竞争%进化算法%Memetic 计算
多目標優化%捕穫競爭%進化算法%Memetic 計算
다목표우화%포획경쟁%진화산법%Memetic 계산
multi-objective optimization%preying-competition%evolutionary algorithm%Memetic computation
本文针对复杂多目标优化问题 Pareto 前沿搜索难度大的特点,设计了一种结合多种群间捕获竞争、强化学习机制的多种群 Memetic 学习策略与进化计算模型.受种群进化、捕食种群与被捕食群体间的竞争等生态学原理的启发,提出了一种基于生态种群捕获竞争模型的多目标 Memetic 优化算法(Multi-Objective Memetic Algorithm based on Ecological Population Preying-competition Model,ECPM-MOMA).ECPM-MOMA 算法设计并运用了捕获竞争、强化学习算子进行全局搜索,在种群进化过程中结合了 Memetic 搜索算子进行局部搜索.理论分析与实验结果表明,本文所提出的算法具有良好的收敛性能和分布特征,生态种群捕获竞争策略与进化计算模型对于解决复杂多目标优化问题是有效的.
本文針對複雜多目標優化問題 Pareto 前沿搜索難度大的特點,設計瞭一種結閤多種群間捕穫競爭、彊化學習機製的多種群 Memetic 學習策略與進化計算模型.受種群進化、捕食種群與被捕食群體間的競爭等生態學原理的啟髮,提齣瞭一種基于生態種群捕穫競爭模型的多目標 Memetic 優化算法(Multi-Objective Memetic Algorithm based on Ecological Population Preying-competition Model,ECPM-MOMA).ECPM-MOMA 算法設計併運用瞭捕穫競爭、彊化學習算子進行全跼搜索,在種群進化過程中結閤瞭 Memetic 搜索算子進行跼部搜索.理論分析與實驗結果錶明,本文所提齣的算法具有良好的收斂性能和分佈特徵,生態種群捕穫競爭策略與進化計算模型對于解決複雜多目標優化問題是有效的.
본문침대복잡다목표우화문제 Pareto 전연수색난도대적특점,설계료일충결합다충군간포획경쟁、강화학습궤제적다충군 Memetic 학습책략여진화계산모형.수충군진화、포식충군여피포식군체간적경쟁등생태학원리적계발,제출료일충기우생태충군포획경쟁모형적다목표 Memetic 우화산법(Multi-Objective Memetic Algorithm based on Ecological Population Preying-competition Model,ECPM-MOMA).ECPM-MOMA 산법설계병운용료포획경쟁、강화학습산자진행전국수색,재충군진화과정중결합료 Memetic 수색산자진행국부수색.이론분석여실험결과표명,본문소제출적산법구유량호적수렴성능화분포특정,생태충군포획경쟁책략여진화계산모형대우해결복잡다목표우화문제시유효적.
Aiming at the difficulty of searching Pareto front for complex multi-objective optimization,a Memetic learning strategy which combines many of population preying-competition mechanism with reinforcement learning mechanism and evolution-ary computation model was designed.Inspired by ecological principle,such as the population evolution and the competition between predator populations and prey populations,a multi-objective Memtic optimization algorithm (multi-objective Memetic algorithm based on ecological population preying-competition model,ECPM-MOMA)was proposed.In ECPM-MOMA,Preying-competition and Reinforcement Learning operator was designed and applied for global search.Memetic search operator was also applied for local search in the population evolution process.Experimental results show that the proposed algorithm has better convergence perfor-mance and distribution characteristics.The ecological preying-competition strategy and evolutionary computation model is effective for solving complex multi-objective optimization problems.