numFeatures = 6;
numTrain = 10;
numTest = 20;
classPriors = [0.5 0.5];
zipfParam = .1; % the bigger this is the more separable the classes are.
for ii =1:numTrain
    [f1 f2] = simZipf(numFeatures, zipfParam);
    if(rand < classPriors(1))
        db(ii,1:numFeatures) = f1;
        
        db(ii,numFeatures+1) = 1;
    else
        db(ii,1:numFeatures) = f2;
       
        db(ii, numFeatures+1) = 2;
    end
end
for jj =1:numTest
    [f1 f2] = simZipf(numFeatures, zipfParam);
    if(rand < classPriors(1))
        testMat(jj,1:numFeatures) = f1;
        testMat(jj,numFeatures+1) = 1;
    else
        testMat(jj,1:numFeatures) = f2
        testMat(jj, numFeatures+1) = 2;
    end
end
% Call Classifiers
centroidParamClassVect = calcCentroidalParametricClass(testMat(:,1:numFeatures), db, classPriors);
for i=1:numTest
    oneNNclassVect(i) = simplekNN(testMat(i,1:numFeatures), db, 1);
    twoNNclassVect(i) = simplekNN(testMat(i,1:numFeatures), db, 2);
    threeNNclassVect(i) = simplekNN(testMat(i,1:numFeatures), db, 3);    
    db1 = db(find(db(:, numFeatures+1)==1),:);
    db2 = db(find(db(:, numFeatures+1)==2),:);    
    centroidClassVect(i) = centroid(testMat(i,1:numFeatures), db1(:, 1:numFeatures), db2(:, 1:numFeatures));
    AvgcentroidClassVect(i) = Avgcentroid(testMat(i,1:numFeatures), db1(:,1:numFeatures), db2(:,1:numFeatures));    
end


testClass = testMat(:, numFeatures+1);
% Calculate error

oneNNError = sum(abs(testClass' - oneNNclassVect));
twoNNError = sum(abs(testClass' - twoNNclassVect));
threeNNError = sum(abs(testClass' - threeNNclassVect));
centroidClassError = sum(abs(testClass' - centroidClassVect));
AvgcentroidClassError = sum(abs(testClass' - AvgcentroidClassVect));
centroidParamClassError = sum(abs(testClass' - centroidParamClassVect));
oneNNError
twoNNError
threeNNError
centroidClassError
AvgcentroidClassError
centroidParamClassError