一种串—并行混合结构的神经网络模型

One Neural Network Model of Serial-parellel Mixed Structure

针对复杂的模式识别问题,提出了一种串—并行混合结构的多种神经网络模型,首先用ART网络对训练集中的样本进行粗分类,以减小训练集的样本规模,然后用多个BP网络对小训练集进行训练。

This article puts forward a neural network model of serial- parellel mixed structure to complex schema identification. First, use ART network to classfy concentrated training samples roughly, it can reduce training collection＇ s sample scale. Then,use some BP network for smaU training collections.