信息优化的模糊神经网络在建模中的应用研究  

A Modeling Approach Based on Fuzzy Neural Network with Information Optimazation to Small Samples

作　　者：张勇[1] 张彤[1] 张家余[1] 陈晓东[1] 丁世坤 

机构地区：[1]哈尔滨电站设备集团公司

出　　处：《哈尔滨工业大学学报》1996年第6期112-115,共4页Journal of Harbin Institute of Technology

摘　　要：提出采用信息聚类－扩散的模糊神经网络（ＩＣＥＦＮ）来对小样本建立模型的理论．利用信息扩散将分类小样本信息扩散至周围信息点，从而解决神经网络在小样本上无法建模的缺陷．同时，还采用了因素程度空间来扩展输入＼输出结点，减少隐结点数，使网络进一步符合实际系统．该方法用于降水量预测，取得了满意的仿真结果．A fuzzy neural network approach on information clustering and extending to small samples modeling has been presented.This new scheme is based on extending each sample information to its surroundings,and therefore the defect about small samples modeling in Neural Network can be overcome.A new space(Degree Factor Space)which extends network input/output nodes and decreases the hidden nodes is adopted to fit the real system.The new method is applied to rainfall prediction and satisfactory simulation results are obtained.

关 键 词：小样本建模 模糊神经网络 信息扩散 信息聚类 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]