What´s wrong with my neural network´s training?

Dear all,

I have this code for training neural network (type newff) and MATLAB is almost all day busy, but nothing happens (no training newff). Does anyone know, where is the mistake?

load ('trenovaci_modely2_stredni')
D = [stredni_tvar{1,:}];
F = [stredni_tvar{2,:}];
net_stredni_MLP = newff(D,F,20);
net.trainparam.epochs = 100;
net = train(net_stredni_MLP,D,F);
D = stredni_tvar{1,7};
J = sim(net_stredni_MLP,D);
save net_stredni_MLP
disp('Neuronová síť byla uložena')
Thank you for your answers.

Answer 1

[Greg Heath]

 net2 = train(net1,...);

 Now there are 2 nets one trained and one untrained. Which do you think is untrained?

 *Thank you for formally accepting this answer*

Greg

[Veronika]

I think, that net1 is untrained. But I really don´t know, I just think...

[Greg Heath]

Yes. net1 is untrained. However, net1 is what you use in the rest of the program. Use

 net1 = train(net1,...);

OK now?

[Veronika]

I tried this:

load ('trenovaci_modely2_stredni')
D = [stredni_tvar{1,:}];
F = [stredni_tvar{2,:}];
net_stredni_MLP = newff(D,F,20);
net_stredni_MLP.trainparam.epochs = 100;
net_stredni_MLP = train(net_stredni_MLP,D,F);
D = stredni_tvar{1,7};
J = sim(net_stredni_MLP,D);
save net_stredni_MLP
disp('Neuronová síť byla uložena')

And still nothing, only "busy" as usual...

[Veronika]

Today I tried another pictures (figures) for training neural network.

This is the code:

   load ('trenovaci_modely4_nejmensi')
    D = [nejmensi_tvar{1,:}];
    F = [nejmensi_tvar{2,:}];
    net_nejmensi_MLP = newff(D,F,20);
    net_nejmensi_MLP.trainparam.epochs = 100;
    net_nejmensi_MLP = train(net_nejmensi_MLP,D,F);
    D = nejmensi_tvar{1,7};
    J = sim(net_nejmensi_MLP,D);
    save net_nejmensi_MLP
    disp('Neuronová síť byla uložena')

And Command Window returns this:

Out of memory. Type HELP MEMORY for your options.

Error in removeconstantrows>reverse (line 262)
    x = zeros(settings.xrows,q);

Error in nnproc.dperf (line 26)
        AA{j+1} = processFcn.reverse(AA{j},processFcn.settings);

Error in C:\Program Files\MATLAB\R2011b\toolbox\nnet\nnutils\+nnprop\jac_s.p>jac_s (line 105)

Error in staticderiv>calc_jacobian (line 222)
  jWB = nnprop.jac_s(net,data.P,data.Pd,data.Zb,data.Zi,data.Zl,...

Error in staticderiv (line 85)
        out1 = calc_jacobian(in2,in3,in4);

Error in defaultderiv>calc_jacobian (line 217)
    jWB = staticderiv('jacobian',net,data,fcns);

Error in nntraining.perfs_jejj>calc_Y_trainPerfJeJJ (line 87)
  Jwb_y = fcns.deriv.calc_jacobian(net,signals,fcns);

Error in nntraining.perfs_jejj>singlecalc (line 58)
  [Y,trainPerfy,trainN,JEy,JJy] = calc_Y_trainPerfJeJJ(net,data,fcns);

Error in nntraining.perfs_jejj (line 9)
    [trainPerfy,trainN,valPerfy,~,testPerfy,~,JEy,JJy] = singlecalc(net,data,fcns);

Error in trainlm>train_network (line 199)
  [perf,vperf,tperf,je,jj,gradient] = nntraining.perfs_jejj(net,data,fcns);

Error in trainlm (line 113)
  [net,tr] = train_network(net,tr,data,fcns,param);

Error in network/train (line 106)
[net,tr] = feval(net.trainFcn,net,X,T,Xi,Ai,EW,net.trainParam);

Error in nejmensi (line 110)
net_nejmensi_MLP = train(net_nejmensi_MLP,D,F;

Do you know, what does it mean? Thank you for answer.

Answer 2

[Greg Heath]

 I answered this in your previous post.

[Veronika]

I know, but still nothing. Only "busy". I tried to do it with nntool function. I set input vector as D, target vector as F and created new network. But when I click on Train Network appeared warning saying:

Inputs and targets have different numbers of samples.

I don´t know what´s wrong...

[Veronika]

Here I attach variables D, F, J and trenovaci_modely2_stredni.

[Greg Heath]

I think you better go back and reread what I wrote.

Answer 3

[Greg Heath]

 load('nntoolMLP.mat')
whos
%   Name          Size         Bytes   Class   
% 
%   D                 1x1         1776  cell               
%   F                 1x1       616560  cell               
%   J                 1x1         2160  cell               
%   stredni_tvar  2x301        1184736  cell       

 d = cell2mat(D);
f = cell2mat(F);
j = cell2mat(J);
s = cell2mat(stredni_tvar);
whos
  Name               Size            Bytes    Class
  D                    1x1             1776    cell                
  F                    1x1           616560    cell                
  J                    1x1             2160    cell
  stredni_tvar         2x301        1184736    cell  
  d                  208x1             1664    double              
  f                  256x301         616448    double              
  j                  256x1             2048    double              
  s                  464x301        1117312    double

Your dimensions are not compatible.

Start by converting to doubles

Then fix the dimensions.

[Veronika]

Ok, when I do this:

load ('trenovaci_modely2_stredni')
s = cell2mat(stredni_tvar);
D = [stredni_tvar{1,:}];
F = [stredni_tvar{2,:}];
d = cell2mat(D);
f = cell2mat(F);
j = cell2mat(J);
net_stredni_MLP = newff(D,F,20);
net_stredni_MLP.trainparam.epochs = 100;
net_stredni_MLP = train(net_stredni_MLP,D,F);
D = stredni_tvar{1,7};
J = sim(net_stredni_MLP,D);
save net_stredni_MLP
disp('Neuronová síť byla uložena')

Command Window returns this:

Cell contents reference from a non-cell array object.

Error in cell2mat (line 43)
cellclass = class(c{1});

Error in stredni (line 110)
d = cell2mat(D);