摘要
通过C++构造出一个五层的BP神经网络,在满足相对误差要求的情况下,实现了指定样本函数的功能。并针对学习效率和权系数修正常数对算法做了改进,有效地加快了收敛速度。最后讨论了当样本函数中Y值非[0,1]区间时,样本的归一化问题。
Constitute a sketch of five layers BP neural network which fulfils the assigned pattern function while the relative error is satisfied.Improves the learning rate and weight coefficient of the arithmetic .The improvements also bring faster convergence speed. At last discuss the normalizati on about pattern when the value of Y is out of .
出处
《微机发展》
2003年第7期93-96,共4页
Microcomputer Development
