【摘要】 为克服标准进化规划算法变异操作的盲目性和易陷入局部最优的问题,提出具有信息指导的自适应退火进化规划算法。算法充分利用目标函数和变量的变化信息,记录下一步的搜索方向,个体的变异方差采用自适应的形式,随进化代数的增加而减小变异幅度,并在新一代种群的生成中采用退火概率的选择方式,既保证了算法的多样性,又可较好地避免算法陷入局部最优解。通过仿真实验表明,该算法收敛速度较快,特别对于变量数目较多的优化问题,更显示出其优越性,具有解决大规模问题的潜力。更多还原

【Abstract】 To overcome the blindness of mutation and permutation in evolutionary programming(EP), an algorithm of self-adaptive evolutionary programming based on searching information is proposed. The information of object function and variables’ variances is used to record the next searching direction. The individual’s mutation tolerance is adaptive with the increase of evolutionary generation number and annealing probability formation is introduced into the selection of EP. So the algorithm guarantees its diversity and may better void local optimum. The simulation shows that the algorithm can converges soon and has potentiality especial for large-scale optimization.更多还原