奇异值分解在重构模糊决策系统规则库中的应用  

Rebuilding PSG Fuzzy Decision-making System Using SVD Method

作　　者：刘斌[1] 何建敏[1] 曹卉宇[1] 

机构地区：[1]东南大学经济管理学院,江苏南京210096

出　　处：《中国管理科学》2005年第3期103-107,共5页Chinese Journal of Management Science

摘　　要：利用SVD技术重构PSG模糊规则库,降低规则的数量。根据矩阵奇异值的性质,适当取舍模糊规则库后件矩阵Ω的奇异值,得到Ω的近似表示^Ω,利用^Ω重新构建规则库,新规则库的输入变量域的维数比原规则库的维数小,从而有效地降低了模糊规则数。仿真结果证明了该方法是有效的。<Abstrcat>The method using SVD to rebuilding PSG fuzzy model is presented in this paper,in new model,the number of fuzzy rules is reduced.In terms of the property of singular values,an approximation of matrix Ω,which is conducted by consequents of fuzzy rules,is obtained by selecting proper singular values of Ω.Using the approximation of Ω,a new rule base is constructed.Generally,the dimension of imput variables in new rule base is smaller than in original one,thus,the size of rule base is reduced effectively.Finally,a numerical example with simulations is given to demonstrate the method discussed throughout this paper.

关 键 词：PSG模糊模型 规则库 奇异值分解(SVD) 约简 

分 类 号：TP273.4[自动化与计算机技术—检测技术与自动化装置] C931[自动化与计算机技术—控制科学与工程]