含二次Sigmoidal神经元的前向网络的分类能力

On the Classification Capability of Feedforward Network Containing Quadratic Sigmoidal Neurons

本文分析了一种新的神经元模型──二次Sigmoidal神经元对前向种经网络的分类能力的改进程度．结果表明：在隐层及输出层无论采用多阈值还是单阈值二次Sigmoidal神经元，三层前向网络的分类能力比CommittedMachine只能提高3倍．

This paper addresses the improvement of the classification capability of network by using quadraic sigmoidal neurons. The results indicate that the degree on the improvement on the classification capability of network by using quadratic sigmoidal neurons is 3.