具有学习行为的协同量子粒子群算法  

Cooperative quantum-behaved particle swarm optimization algorithm with learning behavior

作　　者：董虎胜[1,2] 陆萍[2] 龚声蓉[1] 

机构地区：[1]苏州大学计算机科学与技术学院,江苏苏州215021 [2]苏州经贸职业技术学院信息系,江苏苏州215009

出　　处：《计算机应用研究》2014年第9期2588-2591,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(61170124)

摘　　要：提出了一种具有学习行为的协同量子粒子群算法(LCQPSO)。针对量子粒子群(QPSO)存在的早熟收敛问题,从两方面对其进行改进:引入多子群协同搜索策略提高种群的全局搜索能力,使其在进化后期依然保持多样性;赋予粒子学习行为,提高种群的局部搜索能力。实验中对LCQPSO算法的子群规模与学习概率参数进行了分析,并利用标准测试函数对LCQPSO与PSO、QPSO等算法进行了比较测试,结果表明LCQPSO算法具有更优秀的收敛速度与精度,且能够有效地避免陷入局部极值。This paper proposed a cooperative quantum-behaved particle swarm optimization algorithm with learning behavior （LCQPSO）. For the problem of premature convergence of quantum-behaved particle swarm optimization （QPSO） , the QPSO was improved from 2 ways. First, it adopted the strategy of cooperative searching by muhi-subpopulations to enhance the global searching ability, it also ensured the population＇ s diversity in the later period of evolution. Second, it gave particles the abili- ty of learning from others, which yielded population＇ s better local searching ability. In the experiments, this paper discussed the subpopulation＇ s scale and particle＇ s learning probability parameter, and used several well-known benchmark functions to evaluate the performance of LCQPSO, along with the comparison to PSO, QPSO, etc. The simulation results show that LCQP- SO has faster convergence speed and better precision, it can avoid falling into the local extreme effectively.

关 键 词：量子粒子群 协同进化 学习行为 收敛 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]