文件名称:广义异或集成神经网络算法
- 所属分类:
- 人工智能/神经网络/遗传算法
- 资源属性:
- [Windows] [Visual C] [源码]
- 上传时间:
- 2012-11-26
- 文件大小:
- 49kb
- 下载次数:
- 0次
- 提 供 者:
- 刘*
- 相关连接:
- 无
- 下载说明:
- 别用迅雷下载,失败请重下,重下不扣分!
介绍说明--下载内容均来自于网络,请自行研究使用
本程序用C语言实现了集成神经网络解决广义异或问题。用神经网络集成方法做成表决网,可克服初始权值的影响,对神经网络分类器来说:假设有N个独立的子网,采用绝对多数投票法,再假设每个子网以1-p的概率给出正确结果,且网络之间的错误不相关,则表决系统发生错误的概率为
Perr = ( ) pk(1-p)N-k 当p<1/2时 Perr 随N增大而单调递减.
在工程化设计中,先设计并训练数目较多的子网,然后从中选取少量最佳子网形成表决系统,可以达到任意高的泛化能力。
-this program with C language integrated generalized neural network solutions vary or issues. Using neural network integration methods create networks division, will be able to overcome the initial weights of the neural network classifiers example : Suppose N separate subnets, using an absolute majority vote, Suppose each subnet to 1-p is the probability of correct results, and network between the error is not relevant, The voting system is wrong with a probability of Perr = () pk (1-p) N-k p
Perr = ( ) pk(1-p)N-k 当p<1/2时 Perr 随N增大而单调递减.
在工程化设计中,先设计并训练数目较多的子网,然后从中选取少量最佳子网形成表决系统,可以达到任意高的泛化能力。
-this program with C language integrated generalized neural network solutions vary or issues. Using neural network integration methods create networks division, will be able to overcome the initial weights of the neural network classifiers example : Suppose N separate subnets, using an absolute majority vote, Suppose each subnet to 1-p is the probability of correct results, and network between the error is not relevant, The voting system is wrong with a probability of Perr = () pk (1-p) N-k p
(系统自动生成,下载前可以参看下载内容)
下载文件列表
jc
..\Debug
..\JC.BAK
..\jc.CPP
..\jc.dsp
..\jc.dsw
..\jc.ncb
..\jc.opt
..\jc.plg
..\复件 jc.CPP
..\Debug
..\JC.BAK
..\jc.CPP
..\jc.dsp
..\jc.dsw
..\jc.ncb
..\jc.opt
..\jc.plg
..\复件 jc.CPP