首页 | 本学科首页   官方微博 | 高级检索  
     

BP神经网络结构参数的计算机自动确定
引用本文:刘耦耕,贺素良. BP神经网络结构参数的计算机自动确定[J]. 计算机工程与应用, 2004, 40(13): 72-74,146
作者姓名:刘耦耕  贺素良
作者单位:湖南工业大学电气工程系,湖南,株洲,412008;湖南工业大学电气工程系,湖南,株洲,412008;中南大学信息科学与工程学院,长沙,410083
基金项目:湖南省自然科学基金项目(编号:01jjy3029)
摘    要:研究表明,由多层FNN的BP算法误差函数构成的非线性方程组的独立方程个数和FNN的待求未知变量的个数应该相等,该方程组才能有唯一组解。由此导出网络结构方程式,进而导出隐层层数判别式和每层神经元个数判别式。依据Kolmogorov定理,由该判别式得出求解FNN隐层层数和每个隐层神经元个数的具体算法。计算机仿真结果表明该方法简明实用。

关 键 词:BP神经网络  前向神经网络  隐层层数  神经元个数
文章编号:1002-8331-(2004)13-0072-03

Determining Automatically Structure Parameters of BP Artificial Neural Network by a Computer
Liu Ougeng He Suliang . Determining Automatically Structure Parameters of BP Artificial Neural Network by a Computer[J]. Computer Engineering and Applications, 2004, 40(13): 72-74,146
Authors:Liu Ougeng He Suliang
Affiliation:Liu Ougeng 1 He Suliang 1,21
Abstract:If the equation's quantities of nonliear equation team constituted by error function of BP algorithm of multi-layer feed-forward neural network(FNN)are equal to the one of unknown variables of FNN,the solution of the equations is unique.The network structure equation is then proposed.According to the equation,discriminants of quantity of hidden layers and perceptrons of each layer are given.Depending on Kolmogorov throrem,a particular algorithm of determining quantity of hidden layers and perceptrons of each layer is drawn from the discriminant.The algorithm that is achieved by a computer program is obvious and practical.
Keywords:BP artificial neural network  feed-forward neural network structure  quantities of neurous  quantities of nearous
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号