码迷,mamicode.com
首页 > 其他好文 > 详细

Matlab中常见的神经网络训练函数和学习函数

时间:2017-07-17 21:58:10      阅读:316      评论:0      收藏:0      [点我收藏+]

标签:work   eve   ada   神经元   text   适应   name   调整   record   

一、训练函数

1、traingd

Name:Gradient descent backpropagation (梯度下降反向传播算法 )  

Description:triangd is a network training function that updates weight and bias values  according to gradient descent.              

2、traingda

Name:Gradient descent  with adaptive learning rate backpropagation(自适应学习率的t梯度下降反向传播算法)   

Description:triangd is a network training function that updates weight and bias values  according to gradient descent with adaptive learning rate. it will return a trained net (net) and  the trianing record (tr).

3、traingdx

name:Gradient descent with momentum and adaptive learning rate backpropagation(带动量的梯度下降的自适应学习率的反向传播算法)  

Description:triangdx is a network training function that updates weight and bias values  according to gradient descent momentum and an adaptive learning rate.it will return a trained net (net) and  the trianing record (tr).   

4、trainlm

Name:Levenberg-Marquardt backpropagation (L-M反向传播算法)

Description:triangd is a network training function that updates weight and bias values  according toLevenberg-Marquardt optimization. it will return a trained  net (net) and  the trianing record (tr). 

注:更多的训练算法请用matlab的help命令查看。

 

二、学习函数

1、learngd

Name:Gradient descent weight and bias learning function (梯度下降的权值和阈值学习函数)  

Description:learngd is the gradient descent weight and bias learning function, it will return the weight change dW and a new learning state.

2、learngdm

Name:Gradient descent with momentum weight and bias learning function (带动量的梯度下降的权值和阈值学习函数)  

Description:learngd is the gradient descent  with momentum weight and bias learning function, it will return the weight change dW and a new learning state.

注:更多的学习函数用matlab的help命令查看。

 

三、训练函数与学习函数的区别

  学习函数的输出是权值和阈值的增量,训练函数的输出是训练好的网络和训练记录,在训练过程中训练函数不断调用学习函数修正权值和阈值,通过检测设定的训练步数或性能函数计算出的误差小于设定误差,来结束训练。    
或者这么说:训练函数是全局调整权值和阈值,考虑的是整体误差的最小。学习函数是局部调整权值和阈值,考虑的是单个神经元误差的最小[1]。


参考链接:【1】 https://zhidao.baidu.com/question/1883990061249711708.html?fr=iks&word=matlab%D6%D0traingdx%BA%CDlearngdm%B5%C4%C7%F8%B1%F0&ie=gbk

 

Matlab中常见的神经网络训练函数和学习函数

标签:work   eve   ada   神经元   text   适应   name   调整   record   

原文地址:http://www.cnblogs.com/hezhiyao/p/7197141.html

(0)
(0)
   
举报
评论 一句话评论(0
登录后才能评论!
© 2014 mamicode.com 版权所有  联系我们:gaon5@hotmail.com
迷上了代码!