多层神经网络BP算法的改进

Improvement of Artificial Neural Network with Back Propagation Algorithm

多层神经网络ＢＰ算法的改进姚瑞波孙国雄汤崇熙（东南大学机械工程系，南京２１００１８）目前，前馈型多层神经网络模型已广泛应用于模式识别、语音识别、数据压缩等领域．ＢＰ算法作为其学习方式有效地解决了ＸＯＲ、Ｔ－Ｃ匹配问题，但ＢＰ网络的学习过程是对一个高...

The multilayer feed forward neural network with the BP algorithm is analyzed and a new algorithm to improve the calculating accurary and to increase the convergence rate by using a sine function as the activation function of the neurons is presented. The effectiveness of this method is illustrated by utilizing some computed examples.