摘要
差分进化算法是一种新颖的进化计算技术,为减少用户选择算法控制参数的盲目性和提高算法收敛速度,设计了一种基于并行优进策略的差分进化算法(DEPES算法).算法随着搜索过程的进行随机动态调整缩放因子和选取差分进化模式;在进行差分操作的并行运算过程中,利用当前代最优个体产生新的试验向量参与竞争选择过程.几个复杂函数的数值实验结果表明,DEPES算法寻优效率高、收敛速度快、对初值具有很强的鲁棒性、对维数具有较好的适应性,尤其是具有避免局部极小的能力,其优化性能优于标准DE算法.
Differential evolution(DE) is a new evolutionary computation technology. In order to reduce the difficulty in selecting algorithm parameters and improve convergence, this paper proposed an effective differential evolution based on parallel eugenic strategy ( DEPES). The main principle of DEPES is that parallel eugenic strategy of the new test vector reproduced by the best individual is adopted, and simultaneously, the scale factor and differential strategy adjusted generation by generation at random to avoid the artificial factors. Numerical simulation results on complex functions show that DEPES is effective, efficient, robust to initial conditions and very adaptive to the dimensions, and of excellent ability to avoid being trapped in local minima. Its performances are superior to DE algorithm.
出处
《厦门理工学院学报》
2009年第3期73-78,共6页
Journal of Xiamen University of Technology
关键词
差分进化算法
并行试验向量
优进策略
函数优化
differential evolution
parallel eugenic test vector
eugenic strategy
function optimization