基于极小熵的模糊产生式规则权重获取

加权模糊产生式规则是不确定性知识表示的一种最基本的最常用的形式。一般权重调整的准则是依据于训练精度的提高,但该方法容易引起过度拟合。为了克服传统方法的不足,增强模糊规则的泛化能力,文章提出了使用极小熵定理作为权重调整的新准则,给出了依据极小熵定理求解权重的数学模型,指出了可采用遗传算法求解该数学模型。实验结果表明,训练精度和测试精度随着训练集的模糊熵的减小而增大,同时根据数学模型抽取的权重能够增强模糊规则的泛化能力。

Weight Acquisition of Fuzzy Production Rules Based on Minimum Entropy Principle

Weighted Fuzzy Production Rules(WFPRs) is a fundamental and important way of imprecise knowledge representation.Generally speaking,the usual criterion of the weight values adjustment,which is based only on improving training accuracy,often results in an over-fitting.In order to overcome the shortcoming of over-fitting and improve the generalization capability of WFPRs,this paper proposes a new criterion based on the well-known Minimum Entropy Principle(MEP),presents a mathematics model for acquiring the weight values,and shows a Genetic Algorithm to solve this model.The experimental results show that the training and testing accuracy will increase when the value of fuzzy entropy on the training set decreases,and the weights acquired according to MEP can lead to an enhancement of generalization capability of WFPRs for selected databases.