位置:成果数据库 > 期刊 > 期刊详情页
遗传算法交叉算子的实质分析
  • 期刊名称:北京工业大学学报, 2010, 36(10): 1328-1336 (EI检索)
  • 时间:0
  • 分类:TP301.6[自动化与计算机技术—计算机系统结构;自动化与计算机技术—计算机科学与技术]
  • 作者机构:[1]北京工业大学实验学院,北京100124, [2]北京工业大学电子信息与控制工程学院,北京100124
  • 相关基金:国家自然科学基金资助项目(60704036); 北京市教委科技计划面上资助项目(JC024000200901)
中文摘要:

通过分析比较标准二进制交叉算子和标准十进制交叉算子的异同点,得出结论:交叉算子的实质是在父代个体的数值和所决定的"家族"中随机取值,因而其不能保证交叉操作后的子代个体优于父代个体,体现出盲目搜索的特点;二进制交叉和十进制交叉均对本质交叉搜索区间进行了进一步的划分,但并不能保证进化搜索到更好的样本.根据所得结论,设计了一种改进遗传算法,其有效性证明了本文结论的正确性.

英文摘要:

From the aspect of solution space,the role of crossover operators is analysed firstly. The essence of crossover operators is that it can choose values at random from the solution space included father individuals. So, it is not absolute that the performance of offspring individuals is better than that of father individuals after the father individuals crossed. And it is very easy to bring the aimless search. An improved genetic algorithm based on oriented crossover is proposed. It can make the offspring individuals evolve towards the target value by optimizing their crossover positions and the evolving probability is very large. The simulation results show the algorithm can improve greatly the efficiency and precision in finding the optimum value.

同期刊论文项目
同项目期刊论文