跨层连接的改进BP算法及其应用

Improved BP Algorithm and Application of Cross-layer Link

传统的BP算法存在一些缺陷,它的信息处理能力不仅取决于神经元之间的连接强度,而且与网络的拓扑结构有关。文章将BP传统的顺序连接方式变化为各层神经元之间可以任意连通的跨层连接方式,从而提出了一种改进的BP算法,并将其应用于制茶工业的产品参数设计,实验结果表明,该算法在训练时间、迭代步数和预测效果方面,都远远优于传统的BP算法。

Traditional BP algorithm has some limitations and its information processing capacity depends on connection strength between neurons and network topology as well.A new improved BP algorithm is introduced to change the traditional link way to the cross-layer link between neurons.It is applied to product parameter design in tea production industry.The findings show the new algorithm is far better than the traditional one in training time,iterative steps,prediction results.