Is there anyone familiar with creating a backpropagation neural network using the neural network toolbox found in matlab? any help would be very much appreciated. some examples would be very helpful. Thanks in advance.

 

George,

I have used matlab nnet toolbox to build model for polypropilene properties. If you wish I can send to you some samples of my m-files.

Regards
Leandro

 

Leando,

How you set the parameters for the network in nntool? how to know which performance funcion : MSE, MSEREG and SSE is better for my network?

Regards,
Matilda

 

Matilda,

To set de parameter for your network you need first to create a structure using the command "network". In the following lines I copied a .m files I used to create a model. About your second question. It is not easy to know which is the best function. I have always used "MSE". By my experience, if you are using NN to model a dynamic process, you need to care about the gains of the model, the fitting is not the most important thing.

Regards
Leandro

clear all;
load zuniforme;
x1=XX1;
x2=XX2;
z=ZZ;
[pn,meanp,stdp,tn,meant,stdt]=prestd([x1 x2]',z');

%
net=network; % here you create the structure
%net.userdata
%this parameters are custom parameters I created. To do this edit the .m file netwotk.
net.userdata.stdp=stdp;
net.userdata.meanp=meanp;
net.userdata.stdt=stdt;
net.userdata.meant=meant;
net.userdata.InputMax=max(1.2*[x1 x2],[],1);
net.userdata.InputMin=min(0.8*[x1 x2],[],1);
net.userdata.OutputMax=max(z,[],1);
net.userdata.OutputMin=min(z,[],1);
net.userdata.InputName={'X1';'X2'};
net.userdata.OutputName={'Z'};
net.userdata.note='Z=(X1.^2+X2-11).^2+(X1+X2.^2-7).^2; modelo com Z e X1 uniformes';

%
net.numInputs=2;
net.numLayers=2;
net.biasConnect)=1;
%LAYERS
net.inputConnect=[1 0
1 0]'; %INPUTS

net.outputConnect=[0 1];
net.targetConnect=[0 1];
%segunda camada
net.layerConnect=[0 0
1 0];
%subobjects properties
net.inputs{1}.range=minmax(pn(1,);
net.inputs{2}.range=minmax(pn(2,);

net.layers{1}.size=10;
net.layers{1}.transferFcn='logsig';
net.layers{1}.initFcn='initnw';
net.layers{2}.size=1;
net.layers{2}.transferFcn='purelin';
net.layers{2}.initFcn='initnw';

%network functions
%here you set up the functions.
net.initFcn='initlay';
net.performFcn='mse';
net.trainFcn='trainlm';
%network parameters
net.trainParam.show=1;
net.trainParam.epochs=100;
net.trainParam.goal=1e-3;
if 1
%inicializar a rede
net=init(net);
%treinar a rede
[net,tr]=train(net,pn,tn);
%
%
an=sim(net,pn);
a=poststd(an,meant,stdt);
%
figure
[m,b,r]=postreg(a(1,,z');
end

 

I have the same problem,I want to know how to create a backpropagation neural network using matlab, if you received information that could be helpful to me. I will appreciate your help
Gerhard castro

 

Hi Leandro,

do you you have a tutorial explaining how to use matlab neural network toolbox?. if you may send to me I will be very pleased.

Best Regards.

Coskun
coskun(AT)hamzacebi.net

 

Dear Gerhard
Please visit the following site which may be helpful for you.
http://www-cs.engr.ccny.cuny.edu/~csolb/csc59944/xor.htm
If you have further questions, you can e-mail me at: [email protected]
Best Regards.
Yasser Nabavi

 

hi george, can u somehow send me the code for creating a backpropagation neural network using matlab??? thanx a lot

luckyleaf21(AT)hotmail.com

 

Hello,

I am from Tokyo, i'm young girl, but i have started with A.I. (Artificial Inteligence) in robotics from school.
I am very interested in Japan load forecasting with neural network toolbox and NNTOOL, i have read a lot of theory, papers, etc, but i don't know how start with nntool in forescasting.

Could you help me?,

Thanks,

Yuriko