摘要
针对基本粒子群算法容易早熟及算法震荡问题,提出了一种具有追尾行为的自适应变异粒子群算法,在最优粒子周围添加一个可视区域,如果可视区域内的粒子浓度超过给定标准,则对区域内粒子的个体极值点以一定概率进行自适应变异操作,通过与当前状态比较决定是否更新极值点,变异操作直至粒子离开可视区域、更新了全局极值点或者达到给定变异步数为止。算法增大了搜索能力,而且避免了多余的运算,减少了计算量。通过测试函数仿真验证,结果表明新算法不仅确保收敛、改善了收敛速度,而且有效避免了算法震荡。
In order to avoid the premature convergence and shock of traditional particle swarm optimization algorithm,a novel adaptive mutation algorithm with following behavior is proposed.If paritcles' density in the visual region added around optimal particle is higher than the given criteria, their individual extreme value points will be adaptively mutated in given probability. Then consider whether they are updated when comparing with corrent ones.The mutation operation can't be stopped until it reaches the given mutation step,particles fly away the visual region or update the global extreme value point.The algorithm not only improves searching ability but also reduces computational cost.Simulation with testing function demonstrates that the novel mothod is effective and efficient.
出处
《计算机工程与应用》
CSCD
北大核心
2008年第30期74-76,共3页
Computer Engineering and Applications
基金
国家自然科学基金No.70771079
渭南师范学院人才专项科研基金项目(No.07YkZ007)~~
关键词
粒子群算法
追尾行为
自适应算子
变异算子
Particle Swalan Optimization(PSO)
following behavior
adaptive operator
mutation operator