suddenly and together train by MATLAB

I have studied the artificial neural network models by matlab code as net = newff(PR,[S1 S2...SNl],{TF1 TF2...TFNl},BTF,BLF,PF),

In this code and nntool module in matlab, you can obtain the result for the specified number of neuron such as 2, 3, 4, 10 (default value) in specified neural network architecture such as feed forward, cascade or elman and other.

If you want to compare various numbers of neuron alternatives, you must try one by one to obtain comparative results for artificial neural network structure including various neuron numbers and specified neural network architecture such as feed forward, cascade or elman and other. This process can take a long time for comparison of different artificial neural network models. Is there any matlab code to suddenly and together train and obtain the results for various artificial neural network structures especially various number of neurons, Best regards for helping in this subjets..

Dr. ilker ERCANLI

Answer 1

[ilker ERCANLI]

Dear Greg, especially very much thanks to your response and attention, I produced this flowing Matlab code by examining the references that you said some code including OUTER-LOOP and INNER-LOOP,

for h=1:10; net=newff(dipcap,cap,h,{'logsig'}); net=train(net, dipcap,cap); aa=sim(net,dipcap); end;

ın this code, input variable is dipcap and target varaible is cap, this code resulted in ten number training network including from 1 to 10 number of neurons ranges, but I need predicted cap values of these network models including a various number of neurons, from 1 to 10, especially the code "aa=sim(net,dipcap)" resulted in predictions for last network models including 10 number of neuron, but I need cap predictions from ten various network models includinf from 1 to 10 number of neurons, I used a code "aa(h)=sim(net,dipcap)", but I had a error: "In an assignment A(I) = B, the number of elements in B and I must be the same."

can I solve this problem and can I obtain these cap predictions from ten different network model including from 1 to 10 number of neurons,

best regards...

ilker ERCANLI

[Greg Heath]

Did you realize that you

 1. Wrote in  an ANSWER BOX
 2. Accepted what YOU wrote as the answer

????????????????????????????????????????

I have posted hundreds of dual loop examples in BOTH the NEWSGROUP (comp.soft-sys.matlab) and ANSWERS. Search with a subset of

 greg neural Hmin Hmax Ntrials

Hope this helps.

Greg

Answer 2

[Greg Heath]

I have had very few problems with a double for-loop approach that relies on default parameters except for

   1. OUTER-LOOP over number of hidden nodes h = Hmin:dH:Hmax
   2. INNER-LOOP over random initial weights i = 1:Ntrials

I have posted many, many regression, classification and timeseries examples in BOTH the NEWSGROUP (comp.soft-sys.matlab) and ANSWERS.

Search with one of

 fitnet, patternnet, narxnet, narnet, etc

by including a subset of

 greg, Hmax, Ntrials,

Hope this helps.

Thank you for formally accepting my answer

 Greg