index exceeds matrix dimension

| This is my code i get the error index matrix dimensions in this line: network.meansqrerr {n} = plus( network.meansqrerr {n}, mse( network.error));|

function network = BUILD(inputs, targets, iterations)    

      %     regle pour determiner le noombre des neurones dans la couche caheée
  % source: stackoverflow 
      % Determiner  hidden neuroness ( dans la couche caché couche2) 
  %     round pour les virgules
      nhidden = round((size(inputs, 1) + size(targets, 1)) * 2 / 3);
  % %    
  %     fprintf(' le nombre des couches cachés est: ')
      disp(nhidden) ;
    a=  size(inputs, 1);
      b=  size(inputs, 1);
    disp (a);
    disp(b);
      % Create a neural network 
  %     on determine le nombre des  neurones dans chaque couche
  % couche1:  le nombre des entrées
  % couche 3 (output) le nombre des classe
     network = Create(size(inputs, 1), nhidden, size(targets, 1));

      % Iteration variables

%     SOMME Des erreurs quadratique
    n = 1; 

      % Train network jusk le max des iterations

             %zeros_quat(zeros (10,1));
      while n <= iterations

        network.meansqrerr{n} = [ 0 0 0 0 ];

          % Training epoch
          % nombre des colonnes : size(inputs, 2)
          for i=1:size(inputs, 2)

              % Train network with inputs and target value
              network = traine(network, inputs(:,i), targets(:,i));

%             % Update sum squared error
      network.meansqrerr {n} = plus(  network.meansqrerr {n}, mse( network.error));

   disp(['Iteration: ', num2str(n), ' / '    'Error: ',  num2str(double(network.meansqrerr{n})  )]);
  %          
  % 
  %         % afficher les erreurs dans chaque iteration
  %     disp(['Iteration: ', num2str(n), ' / '    'Error: ', num2str(network.meansqrerr(n) )]);
  % %            

%             

          % Update iteration number
          n = n + 1;

      end
  %     afficher les poids

end

Answer 1

[Walter Roberson]

You initialize

          network.meansqrerr{n} = [ 0 0 0 0 ];

but you iterate

        while n <= iterations

When your n exceeds 4, then network.meansqrerr{n} does not exist in the statement

        network.meansqrerr {n} = plus(  network.meansqrerr {n}, mse( network.error));

By the way: it is confusing that you do

      a=  size(inputs, 1);
      b=  size(inputs, 1);

If you intend both variables to be the same then why not use

      b = a;

??

[AMAL targhi]

sorry but i didn't understand

[AMAL targhi]

     network.meansqrerr{n} = [ 0 0 0 0 ];

 because my meansqrerr is a quaterionic like
network.meansqrer = 0

[AMAL targhi]

i need just to intialise it because my mse(error) is a quaternion so to both must be quaterions

[AMAL targhi]

i had replace network.meansqrerr{n} by network.meansqrerr{1}

but i get the same error

[Walter Roberson]

I cannot proceed any further on this until you explain the difference between this question and your newer http://www.mathworks.com/matlabcentral/answers/288829-index-exceeds-matrix-dimensions-matlab

[Walter Roberson]

Which mse routine are you using?

Please post the complete error message, everything in red, including the part where it says which array the problem is with.

[AMAL targhi]

this is my code

%   network.meansqrerr (n) =     network.meansqrerr (n)+ quattoral( mse( network.error)) ;
%     
% %  disp(['Iteration: ', num2str(n), ' / '    'Error: ',  num2str(double(network.meansqrerr{n})  )]);

function network = BUILD(inputs, targets, iterations)    

      %     regle pour determiner le noombre des neurones dans la couche caheée
  % source: stackoverflow 
      % Determiner  hidden neuroness ( dans la couche caché couche2) 
  %     round pour les virgules
      nhidden = round((size(inputs, 1) + size(targets, 1)) * 2 / 3);
  %    fprintf(' le nombre des couches cachés est: ')
      disp(nhidden)
      % Create a neural network 
  %     on determine le nombre des  neurones dans chaque couche
  % couche1:  le nombre des entrées
  % couche 3 (output) le nombre des classe
      network = Create(size(inputs, 1), nhidden, size(targets, 1));

      % Iteration variables

%     SOMME Des erreurs quadratique
    n = 1; 
%     network.meansqrerr = [];

      % Train network jusk le max des iterations
      while n <= iterations
           network.meansqrerr{n} = [0 0 0 0];
  %        % Training epoch
          % nombre des colonnes : size(inputs, 2)
          for i=1:size(inputs, 2)

              % Train network with inputs and target value
              network = traine(network, inputs(:,i), targets(:,i));
              % Update sum squared error
              %   network.meansqrerr (n) =     network.meansqrerr (n)+ quattoral( mse( network.error)) ;

            network.meansqrerr {n} = times(n^-1,(matric_plus(network.meansqrerr{n},  mse( network.error))))
            size(network.meansqrerr);
          end
          % afficher les erreurs dans chaque iteration

%         disp(['Iteration: ', num2str(n), ' / '    'Error: ', num2str(network.meansqrerr(n) / size(inputs, 2))]);
     disp(['Iteration: ', num2str(n), ' / '    'Error: ',     num2str(cell(network.meansqrerr(n)) / size(inputs, 2))]);

       num2str(double(network.sumsqrerr(n)) / size(inputs, 2))
  % num2str(double(network.sumsqrerr(n)) / size(inputs, 2))

          % Update iteration number
          n = n + 1;

      end
  %     afficher les poids

%     for i = 1:3
%     disp(['Couche' num2str(i)]);
%     for j = 1:1: [1458,59,7]  % Counting neurons in each layer
%         disp(['Neuron ' num2str(i)]);
%         for k = 1:1: nhidden % Counting input to each layer (= # of neurons in the previous layer)
%             disp(['w[' num2str(i) '][' num2str(j) '][' num2str(k) ']=' num2str(network.couches{i}(j,k))]);
%         end
%      
%     end
% end
end

mse is a fucntion for mean square error ( my function)

the error is:

Cell contents reference from a non-cell array object.

Error in matric_plus (line 15) s= plus(a{i,j},b{i,j});

Error in BUILD2 (line 41) network.meansqrerr {n} = times(n^-1,(matric_plus(network.meansqrerr{n}, mse( network.error))))