基于权值与结构确定法的单极Sigmoid神经网络分类器

摘要

构造了以单极Sigmoid函数作为隐层神经元激励函数的神经网络分类器,网络中输入层到隐层的权值和隐层神经元的阈值均为随机生成.同时,结合利用伪逆思想一步计算出隐层和输出层神经元之间连接权值的权值直接确定(WDD)法,进一步提出了具有边增边删和二次删除策略的网络结构自确定法,用来确定神经网络最优权值和结构.数值实验结果表明,该算法能够快速有效地确定单极Sigmoid激励函数神经网络分类器的最优网络结构；分类器的分类性能良好.%A neural network classifier with the hidden neurons activated by unipolar Sigmoid function was constructed and investigated in this paper. The thresholds of hidden neurons and weights between the input layer and the hidden layer of the neural network were randomly generated. The psedoinverse-type Weights Direct Determination (WDD) method was applied to determining the weights between the hidden layer and the output layer. Moreover, a Structure Automatic Determination (SAD) algorithm with pruning-while-growing and twice-pruning policies was proposed to determine the optimal structure of the neural network. The numerical experimental results demonstrate that the SAD algorithm can determine the optimal structure of the neural network quickly and effectively and the neural network classifier has a satisfactory performance.