How do I display red color in true cases for a training data set which has 1000 rand values ?

Here is the program which I am using for generating true cases for green color.

% Define number of nodes in each layer
nLayers     = 3;
nNodesL1    = 3;
nNodesL2    = 3;
nNodesL3    = 1;

% Initialize vectors and matrices
W1 = [ -5 5 -5; -10 20 -10; -5 5 -5 ];
W2 = [ 10  25  10 ];
b1 = [ -5 -1 -5 ]';
b2 = [-5];

% Pack the individual weights and biases into a cell array
arrayW{1} = W1;
arrayW{2} = W2;
arrayB{1} = b1;
arrayB{2} = b2;

% Define the training set
nTr     = 1000;
xTrSet  = rand(3,nTr);
yThresh = 0.75;
yTrSet  = ( xTrSet(2,:)./ ( xTrSet(1,:).^2 + xTrSet(2,:).^2 + xTrSet(3,:).^2 ) .^ 0.5 ) > yThresh;  
imshow(xTrSet(1,:));
% Plot the training set and scores

figure;
hold on

% Specify number of columns for display
nPlotCols = 40;

% Initialize row offset
yPlotOffset = -1;

for iTr = 1 : nTr

    % Check if end of row has been reached
    if mod(iTr-1,nPlotCols) == 0
        yPlotOffset = yPlotOffset + 1;
        xPlotOffset = 0;
    end
    
    % Set corner points for each patch based on row and column indices
    xCorners = xPlotOffset + [ 0 1 1 0 ]';
    yCorners = yPlotOffset + [ 0 0 1 1 ]';
    
    % Set patch colour based on input colour from training case
    patchColor = xTrSet(:,iTr)';
    
    % Draw the patch
    patch(xCorners,yCorners,patchColor);
    
    % Display flag (0 or 1) indicating if the training case passes the
    % "colour test"
    hText = text(xPlotOffset+0.5,yPlotOffset+0.5,sprintf('%d',yTrSet(iTr)>yThresh));
    
    % Set text alignment and size
    set(hText,'HorizontalAlignment','center');
    set(hText,'FontSize',8);
    
    % Set text color to be visible (black on bright, white on dark)
    if norm(patchColor) > 1
        set(hText,'Color',[0 0 0]);
    else
        set(hText,'Color',[1 1 1]);
    end
    
    % Increment column index
    xPlotOffset = xPlotOffset + 1;

end

hold off

title('Training Set')

axis('equal')

% Number of iterations
nIter = 200;

% Initialize arrays
normArray       = zeros(nIter,nLayers-1);
yCorrectFrac    = zeros(nIter,1);
yCorrect        = zeros(nTr,1);

% Peform iterative back-propagation
for iIter = 1 : nIter
    
    fprintf(1,'Iteration %d\n',iIter);
    fprintf(1,'============\n\n');
    
    oldArrayW = arrayW;
    
    % Backpropagate based on the current weights and biases
    [arrayW,arrayB] = BackPropagate(xTrSet,yTrSet,arrayW,arrayB);
    
    % Check the weight and bias deltas 
    for iLayer = 1 : nLayers - 1

        fprintf(1,'Layer %d: \n\n',iLayer);
       
        diffArrayW = arrayW{iLayer} - oldArrayW{iLayer}
        
        normArray(iIter,iLayer) = norm( diffArrayW , 2 );
        
    end
    
    % Check current weights and biases against training set for prediction accuracy 
    for iTr = 1 : nTr
       
        yEst = ForwardPropagate(xTrSet(:,iTr),nLayers,arrayW,arrayB);

        yCorrect(iTr) = ( yTrSet(iTr) > 0.75 ) == ( yEst > 0.5 );
        
    end
    
    % Determine fraction of "correct guesses"
    yCorrectFrac(iIter) = sum(yCorrect) / nTr;
    
    % Print accuracy to screen
    fprintf(1,'\nAccuracy: %4.1f%%\n',100*yCorrectFrac(iIter));
    
end

% Plot the norm of each iterative update to the weights
figure;
plot(normArray ,'.-')
xlabel('Iteration')
ylabel('Delta');
title('Progressive Updates to Weight Matrix Norms')
grid on

% Plot the prediction accuracy iterative update to the weights
figure;
plot(100*yCorrectFrac,'.-')
xlabel('Iteration')
ylabel('Accuracy (%)');
title('Iterative Accuracy')
grid on

% Validation set
nValid = 100;
xValid = rand(nValid,3); 

% Initialize arrays
yValid          = zeros(nValid,1);
indValidTrue    = zeros(nValid,1);
indValidFalse   = zeros(nValid,1);

% Initialize counters
nIndValidTrue  = 0;
nIndValidFalse = 0;

% Cycle through the validation cases and check the output
for iValid = 1 : nValid
   
    [yValid(iValid),arrayZ,arrayA] = ForwardPropagate(xValid(iValid,:)',3,arrayW,arrayB);
    
    if yValid(iValid) > 0.5
        nIndValidTrue = nIndValidTrue + 1;
        indValidTrue(nIndValidTrue) = iValid;
    else
        nIndValidFalse = nIndValidFalse + 1;
        indValidFalse(nIndValidFalse) = iValid;
    end

end


% Plot the "TRUE" cases

figure;
hold on

% Specify number of columns for display
nPlotCols   = 10;

% Initialize row offset
yPlotOffset = -1;

for iInd = 1 : nIndValidTrue

    if mod(iInd-1,nPlotCols) == 0
        yPlotOffset = yPlotOffset + 1;
        xPlotOffset = 0;
    end
    
    xCorners = xPlotOffset + [ 0 1 1 0 ]';
    yCorners = yPlotOffset + [ 0 0 1 1 ]';
    
    patchColor = xValid(indValidTrue(iInd),:);
    
    patch(xCorners,yCorners,patchColor);
      
    xPlotOffset = xPlotOffset + 1;
 
end

hold off

title('TRUE Cases')

axis('equal')


% Plot the "FALSE" cases

figure;
hold on

% Specify number of columns for display
nPlotCols = 10;

% Initialize row offset
yPlotOffset = -1;

for iInd = 1 : nIndValidFalse

    if mod(iInd-1,nPlotCols) == 0
        yPlotOffset = yPlotOffset + 1;
        xPlotOffset = 0;
    end
    
    xCorners = xPlotOffset + [ 0 1 1 0 ]';
    yCorners = yPlotOffset + [ 0 0 1 1 ]';
    
    patchColor = xValid(indValidFalse(iInd),:);
    
    patch(xCorners,yCorners,patchColor);
 
    xPlotOffset = xPlotOffset + 1;

end

hold off

title('FALSE Cases')

axis('equal')

[Image Analyst]

Please format as text so we can get the "Copy" button to copy your code so we can easily paste it into MATLAB and try it.

[Shubhank Sondhiya]

I have attached all the files for your understanding.

Thanks,

Shubhank