基于可生长结构的自组织神经网络研究  

Study of Self-organizing Network Based on Growing Structure

作　　者：孙亮[1] 谢艳辉[1] 于乃功[1] 

机构地区：[1]北京工业大学电子信息与控制工程学院,北京100022

出　　处：《系统仿真学报》2007年第16期3749-3751,共3页Journal of System Simulation

基　　金：国家自然科学基金(60375017);教育部博士学科点专项科研基金(20050005002)

摘　　要：在可生长结构网络(Growing When Required Network,简称GWRN)的基础上,提出了有监督可生长结构网络(Supervised Growing When Required Network,简称SGWRN)模型。该模型引入线性输出层,将GWRN与径向基函数结合,构成有监督自组织神经网络学习模型。该模型能快速生长,可广泛用于监督学习。倒立摆平衡控制仿真实验结果表明该模型有效。A model called Supervised Growing When Required Network （SGWRN） was proposed based on Grow When Required Network （GWRN）. With the idea of adding a linear output layer to GWRN and combining with Radial Basis Function, a supervised self-organizing neural network was formed. This model grows rapidly and can be widely used in supervised learning. Results of simulation to control inverted pendulum show that SGWRN model is available.

关 键 词：自组织 生长网络 拓扑保持性 监督学习 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]