摘要
本文介绍了一种用遗传算法对神经网络和连接权值同时优化的方法。该神经网络的神经元节点可以一定程度地反馈连接,再通过基因链矩阵把神经网络的结构完整地表示。遗传进化学习后,最优个体即是结构最优的神经网络。使用该方法可以设计出结构未知的神经网络。本文最后对XOR问题进行了计算。
A geneiic algorithm to optimize both neural network and related
connection weights in a given task is proposed. A neural network having arbitrary connections is regarded as a virtural living thing which has genes respresenting ita connections among neural units. An XOR problem ia dealt with. The proposed method is available for designing a neural network whose adequate structure ia unknown.
出处
《计算机应用与软件》
CSCD
1997年第3期59-65,共7页
Computer Applications and Software
关键词
神经网络
遗传算法
优化
Neural network, genetic algorithm, optimization.