how to display (show) the similarty of test image in neural network

hi every body.... i used neural network...i want when enter the test image the neural network display the most similarty image of test image...how can do that?? plz help me??

the code which used is :

        function taning2
  load dataset2;

      mynet = newff(P,T,50);

mynet.trainParam.epochs = 3000;
mynet.trainParam.goal        =1e-6;

      mynet.trainParam.lr          = 0.01;

      mynet.divideFcn = 'dividerand';  % Divide data randomly
  mynet.divideMode = 'sample';  % Divide up every sample
  mynet.divideParam.trainRatio = 70/100;
  mynet.divideParam.valRatio = 15/100;
  mynet.divideParam.testRatio = 15/100;
  mynet.trainParam.show        = 100;
  mynet.trainparam.mc          = 0.95;
   mynet.trainParam.max_fail       = 30;

      mynet.trainFcn = 'trainscg';

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

% Train the Network
[mynet,tr] = train(mynet,P,T);

% Test the Network
outputs = mynet(P);
errors = gsubtract(T,outputs);
performance = perform(mynet,T,outputs);

      trainTargets = T.* tr.trainMask{1};
  valTargets = T  .* tr.valMask{1};
  testTargets = T  .* tr.testMask{1};
  trainPerformance = perform(mynet,trainTargets,outputs);
  valPerformance = perform(mynet,valTargets,outputs);
  testPerformance = perform(mynet,testTargets,outputs);

      save mynet

and the files which used to testenter image is :

      function testing2

load mynet;
load dataset2;
image_dims = [46, 64];

images2 = [];
num_images1=1;
m=imread('E:\matlab\project\neuralnetwork\a\img1.jpg');
if num_images1==1
     images2 = zeros(prod(image_dims), num_images1);
end
img2=imresize(m,[46, 64]); 
    images2(:,1) = img2(:);
%     mean_face = mean(images, 2);
    mean_face4 = mean(images2, 1);
     shifted_images2 = images2 - repmat(mean_face4, 1, num_images1 );
     [evectors1,score1,  evalues1] = pcacov(images2');
     num_eigenface1=16;
% % % % % % %      evectors3=evectors1;
evectors3 = evectors1(:, 1:num_eigenface1);
 score3(1,1)=score1(1,1);
 evalues3=evalues1';
  evalues4(1,1)= evalues3(1,1);

    features2 = evectors3' * shifted_images2;
  features4=features2' ;
  [features3,PS2] = mapminmax(features4);
  features3=features3';
  input=[features3;score3;evalues4];
  [input,PS2] = mapminmax(input');
  input=input';

%tt=[1 0;0 1];
% out=mynet11(input);
% figure,plotconfusion(T,out);
simpleclassOutputs2 = sim(mynet,input);
 class = vec2ind(simpleclassOutputs2);
 disp( class );

simpleclassOutputs2 = sim(mynet,input);
 figure,plotconfusion(simpleclassOutputs2,T);

[Image Analyst]

When you say you want to "display the most similarty image" , does that mean you want to know how to display an image or how to create the neural network to determine similarity?

[Image Analyst]

Your edit did not work - formatting is still very messed up. Try reading this and trying to format your code again: http://www.mathworks.com/matlabcentral/answers/13205-tutorial-how-to-format-your-question-with-markup

[primrose khaleed]

thank you image analysist...i want to create nural network to determine simslarty ..then display the similarity image

[Image Analyst]

I can answer the display part. Use imshow(). For the NN part, you'll have to wait for Greg Heath. In the meantime, review his answers here to other people.

[primrose khaleed]

hi greg...plz can you answer my quetion??

Answer 1

[Greg Heath]

I cannot find a MATLAB code for a nearest-neighbor classifier. It looks like you'll have to code your own.

Looking at the source code of NEWPNN might help.

Greg

[Greg Heath]

This may help

>> lookfor knn

knnsearch - Find K nearest neighbors.

ClassificationKNN - K Nearest Neighbors classification

fitcknn - fit KNN classification model

templateKNN - Create a classification KNN template.

[primrose khaleed]

thank you greg

Answer 2

[Greg Heath]

The answer to your question is: If you classify an input using a MLP like patternnet, you have to compare the input with every training vector of that class in order to determine the most similar.

Q: Does that make sense?

A: No

Q: Why not?

A: There are other classifiers that assign classes using a measure similarity. Search

nearest neighbor

Hope this helps