摘要
反向传播算法的最大缺陷之一是容易陷入局部极小点.为此,我们在基因遗传算法(GA)原理的基础上,增加了区域变化这一重要环节,设计出一种新的启发式遗传操作算法,并将它用于多层神经网络的学习过程.实例结果表明用此算法能快速得到最佳权值分布.
One of the main weak-points of back propagation algorithm of multi-layer neural network is that the optimal procedure is easily stacked into the minimal value locally.Based on the genetic algorithm(GA), a new huristic genetic algorithm is presented which includes an important procedure called searching region chane / re-duction andis of higher efficiency as compared with general genetic algorithm.
In this paper, the algorithm is applied in the weights study of multi-layer neural network, and results in fast optimal weights
出处
《计算机工程》
CAS
CSCD
北大核心
1993年第6期22-25,共4页
Computer Engineering
关键词
神经网络
基因遗传算法
GA / multi-layer neural network / optimization
