一种基于神经网络的专家系统的推理机研究

A study of expert system inference engine based on neutal network

传统的推理方式存在推理效率低和冲突消解问题．提出一种新的解决这一问题的方法．即由正向神经网络推理和反向逻辑推理所组成的混合推理系统．利用种经网络所具有的并行性、可学习性、知识可存储性不但可以解决推理冲突，而且还可以利用并行性提高推理的收敛速度和推理效率．为了说明问题，设计了一个实验系统，经实验取得了较理想的效果，并有实用性．

Traditional inference engine has some problems such as inference inefficient and conflic dispel.This paper advance a new method to solve the above problems. This new inference engine consists offorward neural network inference model and backward logical inference mcdel. That is a mixedinference system. The new method shows better results in the practice. The new inference has functionalcharacteristics.