摘要
神经网络规则抽取是神经网络领域的一个重要方向,但是对抽取的规则评估算法却很少。针对这一问题,提出了神经网络抽取规则评估方法。首先证明所有的规则形式都可以统一为区间的形式,然后在区间算法的基础上提出规则评估方法。评估的标准有四个:覆盖性、准确性、矛盾性,以及冗余性。由于规则的矛盾性和冗余性是规则之间的问题,所以该文仅仅研究规则的覆盖性和准确性,提出了覆盖性判断定理,并提出了覆盖性、准确性判断算法。实例证实了该算法的有效性。
Rule extraction from neural networks is one of the most important fields of Neural Network(NN),but the rule evaluation of NN is seldom studied.An algorithm to evaluate the rules extracted from NN was proposed.The fact that all of forms of rules could be unified as intervals was demonstrated.The algorithm of rule evaluation was based on interval algorithm.The criterions of rule extraction conclude the coverage of rules and the veracity of rules.A judging algorithm for coverage of rules and a judging algorithm for veracity of rules were presented respectively.An example was given to validate the proposed algorithm.
出处
《计算机应用》
CSCD
北大核心
2008年第S2期91-93,共3页
journal of Computer Applications
关键词
神经网络
规则抽取
规则评估
覆盖性
准确性
Neural Network(NN)
rule extraction
rule evaluation
coverage
accuracy
