基于模式聚类的多种群模糊遗传算法的研究
|
摘要
针对基本遗传算法优化过程中存在的收敛慢、早熟等问题,提出了一种基于聚类思想的多种群模糊遗传算法(multi population fuzzy genetic algorithm based on clustering theory)。该算法根据传统遗传算法,基于模式聚类的思想采用多种群、模糊动态交叉和变异、保留优越个体策略,利用了模式聚类的思想建立差异化子种群,使选择算子兼顾到种群多样性;设计了模糊控制系统,对交叉算子和变异算子进行动态调控,在保障种群多样性的同时,兼顾优化算法的搜索性能;设立了优越种群以保存每个进化代种群中的最优个体,避免了最优个体的丢失与破坏,此外,优越种群将继续参与优化以产生更为优良的个体。仿真结果表明,所提出的基于聚类思想的多种群模糊遗传算法(CMPFGA)相比于传统遗传算法(genetic algorithm,GA)、多种群遗传算法(multi-population,MPGA)具有更快的收敛速度,能更有效地对于目标函数全局最优解的搜寻,显著提高了算法的性能,极大改善了遗传算法寻优速度缓慢、难以跳出局部最优解等缺点。
To solve the problems of slow convergence and prematurity in the optimization process of basic genetic algorithm,a multi-population fuzzy genetic algorithm based on clustering theory is proposed.Based on the basic genetic algorithm and the idea of pattern clustering,the strategy of multi-population,dynamic crossover,mutation operator and optimal individual preservation is adopted.According to the fitness value of chromosome,the population is divided into different groups by pattern clustering,which makes the selection operator take into account the diversity of population.In control system,the crossover operator and mutation operator are dynamically regulated,the diversity of population is reasonably guaranteed,and the search performance of optimization algorithm is optimized;a superior population is established to preserve the optimal individuals in each evolutionary generation,and the loss and destruction of the optimal individuals are avoided.In addition,the superior population will continue to participate in optimization to produce better results.Which is an excellent population.Simulation results show that the proposed clustering-based multi-population fuzzy genetic algorithm(CMPFGA)has faster convergence speed and larger objective function than genetic algorithm(GA)and multi-population genetic algorithm(MPGA).The probability of searching global optimal solution by number significantly improves the performance of the algorithm,and greatly improves the shortcomings of slow searching speed and difficult to jump out of local optimal solution.
To solve the problems of slow convergence and prematurity in the optimization process of basic genetic algorithm,a multi-population fuzzy genetic algorithm based on clustering theory is proposed.Based on the basic genetic algorithm and the idea of pattern clustering,the strategy of multi-population,dynamic crossover,mutation operator and optimal individual preservation is adopted.According to the fitness value of chromosome,the population is divided into different groups by pattern clustering,which makes the selection operator take into account the diversity of population.In control system,the crossover operator and mutation operator are dynamically regulated,the diversity of population is reasonably guaranteed,and the search performance of optimization algorithm is optimized;a superior population is established to preserve the optimal individuals in each evolutionary generation,and the loss and destruction of the optimal individuals are avoided.In addition,the superior population will continue to participate in optimization to produce better results.Which is an excellent population.Simulation results show that the proposed clustering-based multi-population fuzzy genetic algorithm(CMPFGA)has faster convergence speed and larger objective function than genetic algorithm(GA)and multi-population genetic algorithm(MPGA).The probability of searching global optimal solution by number significantly improves the performance of the algorithm,and greatly improves the shortcomings of slow searching speed and difficult to jump out of local optimal solution.
引文
[1]HOLLAND J H.Adaptation in natural and artificial systems[M].Ann Arbor:University of Michigan Press,1975.
[2]姚明臣.机器学习和神经网络学习中的若干问题研究[D].大连:大连理工大学,2016.
[3]吴浩扬,常炳国,朱长纯,等.基于模拟退火机制的多种群并行遗传算法[J].软件学报,2000,11(3):416-420.
[4]许文杰,刘希玉.基于改进免疫遗传算法的聚类分析研究与应用[J].计算机科学,2008,35(1):240-210.
[5]蓝璐恺,朱建伟,刘伟莉,等.模糊逻辑遗传算法的新方法[J].计算机工程与设计,2008,29(14):3714-3722.
[6]沈于晴,徐立鸿.一种基于模式聚类思想的遗传算法[J].微型电脑应用,2004(4):6-8,64.
[7]张永库,尹灵雪,孙劲光.基于改进的遗传算法的模糊聚类算法[J].智能系统学报,2015,10(4):627-635.
[8]宫尚宝,郭玉翠.基于遗传算法的模糊聚类分析[J].模糊系统与数学,2010,24(6):123-128.
[9]田彩军,孙国正.基于浮点数编码的求解连续函数优化问题的遗传算法[J].武汉交通科技大学学报,1998(4):50-53.
[10]任子武,伞冶.实数遗传算法的改进及性能研究[J].电子学报,2007(2):269-274.
[11]PARK T,RYU K R.A dual population genetic algorithm with evolving diversity[J].Proc.IEEE Congr.Evol.Comput,2007:3516-3522.
[12]王增强,曾碧.遗传算法中交叉算子的配对策略研究[J].汕头大学学报(自然科学版),2005(4):55-58.
[13]朱灿,梁昔明.一种多精英保存策略的遗传算法[J].计算机应用,2008(4):939-941.
[14]王凯.飞行器位姿光电测量的算法性能评估[D].成都:电子科技大学,2018.
[15]王泰华,贾玉婷.不确定机器人自适应鲁棒迭代学习控制研究[J].软件导刊,2018,17(3):32-36,55.
[2]姚明臣.机器学习和神经网络学习中的若干问题研究[D].大连:大连理工大学,2016.
[3]吴浩扬,常炳国,朱长纯,等.基于模拟退火机制的多种群并行遗传算法[J].软件学报,2000,11(3):416-420.
[4]许文杰,刘希玉.基于改进免疫遗传算法的聚类分析研究与应用[J].计算机科学,2008,35(1):240-210.
[5]蓝璐恺,朱建伟,刘伟莉,等.模糊逻辑遗传算法的新方法[J].计算机工程与设计,2008,29(14):3714-3722.
[6]沈于晴,徐立鸿.一种基于模式聚类思想的遗传算法[J].微型电脑应用,2004(4):6-8,64.
[7]张永库,尹灵雪,孙劲光.基于改进的遗传算法的模糊聚类算法[J].智能系统学报,2015,10(4):627-635.
[8]宫尚宝,郭玉翠.基于遗传算法的模糊聚类分析[J].模糊系统与数学,2010,24(6):123-128.
[9]田彩军,孙国正.基于浮点数编码的求解连续函数优化问题的遗传算法[J].武汉交通科技大学学报,1998(4):50-53.
[10]任子武,伞冶.实数遗传算法的改进及性能研究[J].电子学报,2007(2):269-274.
[11]PARK T,RYU K R.A dual population genetic algorithm with evolving diversity[J].Proc.IEEE Congr.Evol.Comput,2007:3516-3522.
[12]王增强,曾碧.遗传算法中交叉算子的配对策略研究[J].汕头大学学报(自然科学版),2005(4):55-58.
[13]朱灿,梁昔明.一种多精英保存策略的遗传算法[J].计算机应用,2008(4):939-941.
[14]王凯.飞行器位姿光电测量的算法性能评估[D].成都:电子科技大学,2018.
[15]王泰华,贾玉婷.不确定机器人自适应鲁棒迭代学习控制研究[J].软件导刊,2018,17(3):32-36,55.