自底向上粗集神经网络建模方法  

A Bottom-up Approach for Rough Neural Network Modeling

作　　者：张东波[1] 

机构地区：[1]湘潭大学信息工程学院,湖南湘潭411105

出　　处：《湘潭大学自然科学学报》2009年第2期130-138,共9页Natural Science Journal of Xiangtan University

基　　金：湖南省教育厅资助项目(05C093)

摘　　要：提出了一种基于自底向上方式构造模糊粗糙数据模型并实现粗集神经网络建模的方法,该方法通过自适应G-K聚类算法,实现输入论域空间的模糊划分,在基于聚类数和约简属性搜索的基础上,提取优化的模糊粗糙数据模型,在此基础上融合神经网络实现粗集神经网络建模.对Brodatz纹理图像的实验表明,该方法性能优于传统的贝叶斯和LVQ方法,和传统的粗逻辑神经网络RLNN相比,该方法建立的神经网络结构精简,收敛速度快,具有更强的泛化能力.An approach, which based on the fuzzy rough model constructed by a bottom-up modeling mode, to build rough neural network is proposed. By means of adaptive Gaustafason-Kessel （G-K） algo- rithm, fuzzy partition can be implemented in input data space. Based on the search of cluster number and feature reduction, optimization fuzzy rough model can be extracted and rough neural network model will be constructed by integrating neural network technique. The experiment results of classifying Brodatz texture image indicate that the method is superior to conventional Bayesian and LVQ methods, and the rough neural network based on fuzzy rough model has superiorities in construction, convergence speed and generalization ability, compared with traditional rough logic neural network （RLNN）.

关 键 词：粗糙集 粗集神经网络 粗糙集数据分析 纹理分类 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]