| 用最优成员选择方法训练人工神经网络 |
| |
| 引用本文： | 吕庆章 王晓钰. 用最优成员选择方法训练人工神经网络[J]. 计算机与应用化学, 1998, 15(2): 100-103 |
| |
| 作者姓名： | 吕庆章 王晓钰 |
| |
| 作者单位： | 河南师范大学化学系!新乡，453002(吕庆章，曹益林，杨书廷)，河南师范大学化学系!新张师范专科学校化学系，新乡，453002(王晓钰)，河南师范大学化学系!南阳师范专科学校化学系，新乡，453002(陈文涛) |
| |
| 摘 要： | 对遗传算法采用实数编码用于训练神经网络进行了讨论，由实际运算发现，因为只有好的成员才能繁衍，所以无论淘汰率较小或较大时，经过若干代的进化后，往往形成近亲繁衍生息的情况，近亲繁衍时交叉操作对进化作用不大，如何工只取一个最优成员采用突变方法产生子代成员来训练神经网络能得到很好的结果，最翁
|
| 关 键 词： | 人工神经网络 遗传算法 实数编码 突变几率 |
AN OPTIMUM SELECTION SCHEME FOR ARTIFICIAL NEURAL NETWORK TRAINING |
| |
| LU Qing-zhang,WANG Xiao-yu,CHEN Wen-tao,CAO Yi-lin,YANG Shu-ting. AN OPTIMUM SELECTION SCHEME FOR ARTIFICIAL NEURAL NETWORK TRAINING[J]. Computers and Applied Chemistry, 1998, 15(2): 100-103 |
| |
| Authors: | LU Qing-zhang WANG Xiao-yu CHEN Wen-tao CAO Yi-lin YANG Shu-ting |
| |
| Abstract: | The genetic algorithm is applied in the back-propagation neural network training. It is found that inbreed always happens during evolution with real numbers used as the components of chromosome. In this case , the crossover operation for reproduction is not effective in evolution. An optimum selection scheme is proposed in this paper, in which only the optimum member is selected for reproducing the next generation in the mutation operation. It is found that the optimum selection scheme can be applied in the back-propagation neural network training, and the oscillation phenomenon can be avoided when the selected member is taken as one of the children. |
| |
| Keywords: | Artificial neural network Genetic algorithm Real number string Mutation probability Mutation factor |
| 本文献已被 CNKI 维普 等数据库收录！ |
