bp神经网络matlab实例(bp神经网络matlab实例)
- 格式：doc
- 大小：25.10 KB
- 文档页数：23
bp神经网络matlab实例（bp神经网络matlab实例）
Case 1 training BP network by momentum gradient descent
algorithm.
Training samples are defined as follows:
Input vector as
P =[-1 -2 31
-1 15 -3]
The target vector is t = [-1 -1 1 1]
Solution: the MATLAB program of this example is as follows:
Close all
Clear
Echo on
CLC
% NEWFF - generating a new feedforward neural network
% TRAIN -- training BP neural network
% SIM -- Simulation of BP neural network
Pause
Start by hitting any key
CLC
Percent defines training samples
% P as input vector
P=[-1, -2, 3, 1; -1, 1, 5, -3];
% T is the target vector
T=[-1, -1, 1, 1];
Pause;
CLC
% create a new feedforward neural network
Net=newff (minmax (P), [3,1],
{'tansig','purelin'},'traingdm')
The current input layer weights and thresholds
InputWeights=net.IW{1,1}
Inputbias=net.b{1}
The current network layer weights and thresholds
LayerWeights=net.LW{2,1}
Layerbias=net.b{2}
Pause
CLC
% set training parameters
Net.trainParam.show = 50;
Net.trainParam.lr = 0.05;
Net.trainParam.mc = 0.9;
Net.trainParam.epochs = 1000;
Net.trainParam.goal = 1e-3;
Pause
CLC
% call TRAINGDM algorithm to train BP network
[net, tr]=train (net, P, T);
Pause
CLC
Simulation of BP network by%
A = sim (net, P)
Calculate the simulation error
E = T - A
MSE=mse (E)
Pause
CLC
Echo off
Example 2 adopts Bayesian regularization algorithm to improve
the generalization ability of BP network. In this case, we used
two kinds of training methods, namely L-M algorithm (trainlm)
and the Bias regularization algorithm (trainbr), is used to
train the BP network, so that it can fit attached to a white
noise sine sample data. Among them, the sample data can be
generated as follows MATLAB statements:
Input vector: P = [-1:0.05:1];
Target vector: randn ('seed', 78341223);
T = sin (2*pi*P) +0.1*randn (size (P));
Solution: the MATLAB program of this example is as follows:
Close all
Clear
Echo on
CLC
% NEWFF - generating a new feedforward neural network
% TRAIN -- training BP neural network
% SIM -- Simulation of BP neural network
Pause
Start by hitting any key
CLC
% define training sample vector
% P as input vector
P = [-1:0.05:1];
% T is the target vector
Randn ('seed', 78341223); T = sin (2*pi*P) +0.1*randn (size
(P));
Draw the sample data points
Plot (P, T, +);
Echo off
Hold on;
Plot (P, sin (2*pi*P), ':');
Draw sine curves without noise
Echo on
CLC
Pause
CLC
% create a new feedforward neural network
Net=newff (minmax (P), [20,1], {'tansig','purelin'});
Pause
CLC
Echo off
CLC
Disp ('1. L-M optimization algorithm TRAINLM'); disp ('2.
Bayesian regularization algorithm TRAINBR');
Choice=input (\ "please select training algorithm (1,2): ');
Figure (GCF);
If (choice==1)
Echo on
CLC
% using L-M optimization algorithm TRAINLM
Net.trainFcn='trainlm';
Pause
CLC
% set training parameters
net.trainparam.epochs = 500；
net.trainparam.goal = 1e-6；
NET（.NET）；
%重新初始化
暂停
中图分类号
“（选择= = 2）
回声
中图分类号
%采用贝叶斯正则化算法trainbr
trainfcn = 'trainbr”网；
暂停
中图分类号
%设置训练参数
net.trainparam.epochs = 500；
randn（“种子”，192736547）；
NET（.NET）；
%重新初始化
