How can I design my own fitting neural network?

I want to design a neural network which will be trained to calculate the squra root of 1000 random numbers generated. but I don't want to use fitnet, instead I want to design my own fitting network. can any one tell me the architecture of fitnet used in matlab means how many layers, initializing functions, training functions etc used.

Answer 1

[Greg Heath]

Always start with the default configuration.

 help fitnet

Combine with the script obtained via the nntool GUI.

You also can find examples by searching

 greg fitnet Ntrials

HTH

Thank you for formally accepting my answer.

Greg

[saima zia]

Thanks for guiding. I m sending u my code, I have tried to create my own fitnet to train it to calculate square root of number,instead of using the matlab fitnet. the structure of my network is exactly like the fitnet architecture( I analyzed the fitnet architecture by looking at the script and object created of fitnet).plz check it and let me know what mistake I m making cause its not getting trained.

x=rand(1,1000)*150;

t=sqrt(x);

net=network();

net.adaptFcn='adaptwb';

net.numInputs=1;

net.inputs{1}.size=1;

net.numLayers=2;

net.layers{1}.size=10;

net.layers{2}.size=1;

net.inputConnect(1,1)=1;

net.layerConnect(2,1)=1;

net.biasConnect(1)=1;

net.biasConnect(2)=1;

net.biases{1}.learnFcn='learngdm';

net.biases{1}.initFcn='';

net.biases{2}.learnFcn='learngdm';

net.biases{2}.initFcn='';

net.outputConnect(2)=1;

net.layers{1}.transferFcn='tansig';

net.layers{2}.transferFcn = 'purelin';

net.inputweights{1,1}.initFcn='';

net.inputweights{1,1}.learnFcn='learngdm';

net.layerweights{2,1}.initFcn='';

net.layerweights{2,1}.learnFcn='learngdm';

net.inputs{1}.processFcns = {'removeconstantrows','mapminmax'};

net.outputs{2}.processFcns = {'removeconstantrows','mapminmax'};

net.divideFcn = 'dividerand'; % Divide data randomly

net.divideMode = 'sample'; % Divide up every sample

net.divideParam.trainRatio = 70/100;

net.divideParam.valRatio = 15/100;

net.divideParam.testRatio = 15/100;

net.trainFcn = 'trainlm';

net.performFcn = 'mse';

net.plotFcns = {'plotperform','plottrainstate','ploterrhist', ... 'plotregression', 'plotfit'};

[net,tr] = train(net,x,t);

outputs = net(x);

test=100;

result = sim(net,test);