function [k] = findBestK (trainMat)
% trainMat = n*m+1: n objects of dimension m; last column is class
% returns the value of k that returns the best value of kNN

numObjects = size(trainMat,1);
numFeatures = size(trainMat,2)-1;
% trying k= 1 to numObjects-1
numCorrect = zeros(1, numObjects-1);

for i=1:numObjects-1
    for j=1:numObjects
        % k = i, try on each object j
        remTrainMat = zeros(numObjects-1,numFeatures+1);
        if(j==1)
            remTrainMat = trainMat(j+1:numObjects,:);
        elseif(j==numObjects)
            remTrainMat = trainMat(1:j-1,:);
        else
            remTrainMat(1:j-1,:) = trainMat(1:j-1,:);
            remTrainMat(j:numObjects-1,:) = trainMat(j+1:numObjects,:);
        end
        result = simplekNN(trainMat(j,1:numFeatures), remTrainMat, i);
        if(result == trainMat(j,numFeatures+1))
            numCorrect(1,i) = numCorrect(1,i) + 1;
        end
    end
end
[who where] = max(numCorrect);
numCorrect
k = where;