高斯型模糊前向神经网络  

Gaussian Fuzzy Feed-forward Neural Network

作　　者：杨文光[1] 许璐[1] 闫守峰[1] 

机构地区：[1]华北科技学院基础部,北京燕郊101601

出　　处：《模糊系统与数学》2014年第4期120-125,共6页Fuzzy Systems and Mathematics

基　　金：中央高校基本科研业务费资助项目(3142013021);华北科技学院高等教育科学研究课题(HKJYZD201213)

摘　　要：利用高斯型隶属函数和采样数据得到了三层模糊前向神经网络。该网络模型利用权值直接确定法得到了最优权值,并依据采样数据中的插值样本较好确定了单隐层神经元个数。该网络是近似插值神经网络。仿真实验表明,高斯型模糊前向神经网络具有逼近精度高、网络结构简单、良好的去噪性和实时性高等优点。Three-layer fuzzy feed-forward neural network is obtained through the Gaussian memhership function and sampling data. For this model, the best weight is received with the method of weights-direct- determination, and the single hidden layer neurons can be determined from the sampled data interpolation samples. The network is approximate interpolation neural network. Simulation results show that, the Gaussian fuzzy feed-forward neural network has many advantages, such as high precision, simple structure, good denoising and real-time.

关 键 词：前向神经网络 权值直接确定法 高斯型隶属函数 实时 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]