一种适用于模式分类的模糊粗隶属函数神经网络

在设计模式分类器时,由于很难甚至不可能获得实现精确分类的所有必要的本质特征属性,使得在进行分类时出现模糊不确定性和粗糙不确定性并存的情况.基于模糊粗糙集理论,本文构建了一种模糊粗隶属函数神经网络FRMFN.该网络融合了模糊信息和粗糙信息的处理能力,在对加拿大Norman Wells地区的红外波段合成图像进行分类的测试中,显示FRMFN网络具有比相应RBF网络更好的分类精度,同时保留了RBF网络学习速度快的优点.

Fuzzy-Rough Membership Function Neural Network Suitable for Pattern Classification

In the design of a pattern classifier,it is difficult or impossible to acquire all the necessary essential attributes for precise classification,so both the fuzzy uncertainty and rough uncertainty coexist in classification. A FRMFN(Fuzzy-Rough Membership Function Neural Network) was designed,based on fuzzy-rough set theory.The FRMFN integrated the ability to process fuzzy and rough information.The test results of classification for infrared band combination image of Canada Norman Wells territories indicated that the FRMFN had better classification precision than RBF network and had the same merit of quick learning as RBF network.