摘要
基于排序的选择方式在一定程度上会导致种群搜索范围变窄,进化规划算法过早收敛。针对此问题,将退火概率与适应度结合的选择方式引入进化规划算法的选择操作,形成了退火进化规划算法(AEP)。然后利用非时齐Markov链对退火进化规划算法进行了描述,并证明了其全局收敛性。数值实验表明,退火进化规划算法能保证种群的全局收敛性,且收敛速度较快,可较好地避免早熟收敛和局部极值。
Selecting pattern based on ranking in Evolutionary Programming Algorithm causes the population＇s searching space to become narrower and narrower to some degree, which will result in premature convergence. To this problem, annealing probability formation was introduced into the selection of EP and a new algorithm： annealing EP （AEP） was got. Then the time-dependent Markov chains were described and the convergence of AEP was proved It is proved that the algorithm always converges globally and the numerical test shows that AEP can converge soon and it can better avoid premature convergence and local optimum.
出处
《系统仿真学报》
EI
CAS
CSCD
北大核心
2006年第3期577-580,共4页
Journal of System Simulation
