摘要： 本文从神经网络与传统专家系统相结合的角度出发,提出的一种基于层次与或图的神经网络专家系统模型.BP网被用作规则记忆器.并通过源于层次与或图的二值化样本的训练实现知识的获取,而网络的多次正向传播式的应用则用来完成相应的层次推理工作.＊

关键词: 神经网络, 专家系统, 层次结构

Abstract: This paper deals with a neural network expert model based on hierarchical "and" and "or" graphic. The multi-layer feedforward network oriented from back-propagation algorichm is used for realizing the rule memory. Consistent with the above principles, samples composed of binary numbers arc utilized to train the cor-respondent network in order to acquire knowledges, while the rcpcatable forword-propagation of the network is related to the reasoning operations.