how to store target into array bpnn?

if true
  % code
endhi, i'm working on project of classification images. i have 4 classes. the dataset of the training in the last column is the class label where 4 class. 1-4.

how to write a code to store the target into array. i want to make if the target is 4, it will store 1000. my code below did not work.

loadinput = dlmread('Train.csv');<

x = loadinput;

tempt = loadinput;

x(:,2501)=[];

noofInputSet = length(x(:,1));

disp(noofInputSet);

noofInputNode = length(x(1,:));

noofHiddenNode = noofInputNode / 2;

noofInputBiasNode = 1;

noofHiddenBiasNode = 1;

noofOutputNode = 4;

for r=1:noofInputNode

    tempt(:,1)=[];

%store target into array t

for sT=1:noofInputSet

   if tempt(sT)==1,

        t(sT) =1000;

   elseif tempt(sT)==2,

        t(sT) = 0100; 

   elseif tempt(sT)==3,

       t(sT)=0010;

   elseif tempt (sT)==4,

       t(sT)=0001;

   end

-can somebody help me fix this

Answer 1

[Greg Heath]

 TARGET VECTORS SHOULD BE COLUMNS OF THE UNIT MATRIX eye(4)

 classindices = [ 1 2 3 4 3 2 1] 
 target       = full(ind2vec(classindices))
 classindices = vec2ind(target)

 Hope this helps.

Thank you for formally accepting my answer

Greg

[ameer zamri]

thank you greg for your answer,,

my code below need 4 class. i dont know how to set the target and get the Mean Square Error. i'm done trying but when i try...the MSE is not change. it just static. i hope you can help me. below is my full code for bpnn training

loadinput = dlmread('BPNNDataTraining.csv'); x = loadinput; tempt = loadinput; x(:,2501)=[];

noofInputSet = length(x(:,1)); disp(noofInputSet); noofInputNode = length(x(1,:)); noofHiddenNode = noofInputNode / 2; noofInputBiasNode = 1; noofHiddenBiasNode = 1; noofOutputNode = 4;

for r=1:2500 tempt(:,1)=[]; end %store target into array t classtarget =[1 2 3 4]; t = full(ind2vec(classtarget)); classtarget = vec2ind(target)

%step 1: Initialize the weight and bias

v=rand(noofInputNode,noofHiddenNode); %32x16 matrix v1=zeros(noofInputNode,noofHiddenNode); %32x16 matrix for momentum purpose b1=rand(noofInputBiasNode,noofHiddenNode); %1x16 array w=rand(noofOutputNode,noofHiddenNode); %1x16 array w1=zeros(noofOutputNode,noofHiddenNode);%1x16 array for momentum purpose b2=rand(noofHiddenBiasNode,noofOutputNode);

%step 3: For each training pair SelectedPopupLeaningRate = get(handles.LearningRatePopupmenu, 'Value'); SelectedPopupmenuMomentumFactor = get(handles.MomentumFactorPopupmenu, 'Value'); SelectedPopupmenuEpoch = get(handles.EpochPopupmenu, 'Value');

switch SelectedPopupLeaningRate case 1 alpha = 0.1; case 2 alpha = 0.2; case 3 alpha = 0.3; case 4 alpha = 0.4; case 5 alpha = 0.5; case 6 alpha = 0.6; case 7 alpha = 0.7; case 8 alpha = 0.8; case 9 alpha = 0.9; end

switch SelectedPopupmenuMomentumFactor case 1 mf = 0.0; case 2 mf = 0.1; case 3 mf = 0.2; case 4 mf = 0.3; case 5 mf = 0.4; case 6 mf = 0.5; case 7 mf = 0.6; case 8 mf = 0.7; case 9 mf = 0.8; case 10 mf = 0.9;

end

switch SelectedPopupmenuEpoch case 1 finalepoch = 50; case 2 finalepoch = 100; case 3 finalepoch = 500; case 4 finalepoch = 1000; case 5 finalepoch = 2000; case 6 finalepoch = 4000; end

con=1; epoch=0;

PlotGraph = plot(handles.RMSEGraph,0); set(PlotGraph,'Color','blue','LineWidth',2) xlabel('Number of Epoch') ylabel('Root Mean Square Error (RMSE)')

while con e1=0; for I=1:noofInputSet, %input data

        %Feed forward stage
        %step 4: Calculate Z_in
        for j=1:noofHiddenNode
            zin(j)=b1(j);
            for i=1:noofInputNode
                zin(j)=zin(j)+x(I,i)*v(i,j);
            end 
           %calculate the output signal
            z(j)=binarysigmoid(zin(j));
        end
        %Step 5: Calculate Y_in
        for k=1:noofOutputNode
           yin(k)=b2(k);      
           for j=1:noofHiddenNode
                yin(k)=yin(k)+z(j)*w(k,j);         
           end  
           %calculate the output signal
           y(k)=binarysigmoid(yin(k));
           %ty(I,k)=y(k);
        end

        %backpropagationtraining of Error
        %step 6: Calculate the output
        for k=1:noofOutputNode %lowercase delta k
            delk(k)=(t(I)-y(k))*binarysigmoid1(yin(k));
        end

        %step 7
        for j=1:noofHiddenNode %lowercase delta inj
           for k=1:noofOutputNode
               delinj(j)=delk(k)*w(k,j);
           end
        end 

        for j=1:noofHiddenNode%lowercase delta j
            delj(j)=delinj(j)*binarysigmoid1(zin(j));
        end 

        %Weight updation
        %step 8
        for i=1:noofInputNode%uppercase delta v(i,j)
            for j=1:noofHiddenNode
                delv(i,j)=alpha*delj(j)*x(I,i)+mf*(v(i,j)-v1(i,j));
            end
        end
        delb1=alpha*delj; 
        v=v+delv;
        b1=b1+delb1;

        for j=1:noofHiddenNode %uppercase delta w(j,k)
           for k=1:noofOutputNode
                delw(k,j)=alpha*delk(k)*z(j)+mf*(w(k,j)-w1(k,j));  
           end
        end    
        delb2=alpha*delk; %uppercase delta bias2(j,k)
        w=w+delw;
        b2=b2+delb2;

        w1=w; %store in current w to w1 for momentum purpose
        v1=v; %store in current v to v1 for momentum purpose

        for k=1:k
            mse = ((t(I)-y(k))^2);
            e1=e1+mse;
        end

    end
    e=sqrt(e1/noofInputSet);
    disp(e);
    %step 9
    %test stopping condition
    if e<0.005
        con=0;
    end
    epoch=epoch+1;
    graphMSE(epoch) = e;
    if epoch==finalepoch;
        con=0;
    end
    PlotGraph = plot(handles.RMSEGraph,graphMSE);
    set(handles.uitable2,'Data',v);
    set(handles.uitable3,'Data',b1);
    set(handles.uitable4,'Data',w);
    set(handles.uitable5,'Data',b2);
    set(PlotGraph,'Color','blue','LineWidth',2)
    xlabel('Number of Epoch')
    ylabel('Root Mean Square Error (RMSE)')
    stre = num2str(e);
    set(handles.MeanSquareErrorText2,'string',stre);
    strepoch = num2str(epoch);
    set(handles.CurrentEpochText2,'string',strepoch);
    pause(.0001);

end

[Greg Heath]

You probably won't get any help until you correctly format your code.

Greg