用前向神经网络获取特定结构的规则知识

介绍一种用前向神经网络来获取特定结构的产生式规则知识的方法，这些规则知识既可以用于解释神经网络的行为，又可以用于问题求解，获取的产生式规则知识可以是任意形式的，也可以是包含多个推理步的多级推理链知识，而且具有所需的易于理解的特一表达结构，另外，文中还提出了解决智能系统形成过程中知识不断增长问题的方法，大量的逻辑表达式学习实验结果表明，文中获取规则知识的方法是有效的。

SPECIALLY STRUCTURED RULE KNOWLEDGE ACQUISITION WITH FEED FORWARD NEURAL NETWORKS

A method to acquire production rule knowledge with special structure by use of a feed forward neural network(FNN) is presented. The rule knowledge acquired can be used to explain the result obtained by FNN and to solve some problems. With the propoed method, production rule with any form can be acquired and inference link knowledge with multi inference steps can be acquired as well. And these production rules can be expressed with the special structure and form expected by man.A method to solve the problem of knowledge growing in some intelligent systems is presented. Many simulation experiments to learn logical expressions are proformed with this method and the validity of this method is proved by our experiment results.