crossValidation(model, m_dap)

function malclassificationList = crossValidation(model, m_dap)

% Cross-Validation

% MacLachlan, G.J. Discriminant Analysis and Statistical Pattern
% Recognition, Wiley

malclassificationList = cell([1, m_dap.constants.numGrp]);

disp(['Starting Cross-Validation ', datestr(now)])

for i = 1:m_dap.constants.numGrp
    for j = 1:m_dap.constants.numElemPerGrp(i)
        dR = []; % Discriminant Rule
        dataElem = m_dap.datasets.dataPerGroups{i}(j,:);
        if strcmp(model, 'linear')
            
%             invCov = m_dap.modeledInvCovList.linear{1};

            tempCovList = m_dap.covList;
            tempCovList{i} = cov(m_dap.datasets.dataPerGroups{i}([1:j-1 j+1:end],:));
            pooledCov = computeLinearModel(tempCovList, m_dap.constants);
            invCov = inv(pooledCov);

            for k = 1:m_dap.constants.numGrp
                
%                 cvCoef = linearCVCoefficent(invCov, dataElem, k, i, m_dap); %TODO: check Cross Validation (leave-one-out) functionality

                cvCoef = 1;
                dR(k) = linearRule(m_dap.meanList{k}, invCov, dataElem, m_dap.constants.aprioriProb(k), cvCoef);
            end
        elseif strcmp(model, 'cpc')
            tempCovList = m_dap.covList;
            tempCovList{i} = cov(m_dap.datasets.dataPerGroups{i}([1:j-1 j+1:end],:));
            cpcCovList = computeCPCModel(tempCovList, m_dap.constants);
            tempInvCovList = computeInvCovList(cpcCovList, m_dap.constants.numGrp);
            cvCoef.D = 1;
            cvCoef.detD = 1;
            for k = 1:m_dap.constants.numGrp
                dR(k) = quadraticRule(m_dap.meanList{k}, tempInvCovList{k}, dataElem, m_dap.constants.aprioriProb(k), cvCoef);
            end
        elseif strcmp(model, 'proportional')
            tempCovList = m_dap.covList;
            tempCovList{i} = cov(m_dap.datasets.dataPerGroups{i}([1:j-1 j+1:end],:));
            proportionalCovList = computeProportionalModel(tempCovList, m_dap.constants);
            tempInvCovList = computeInvCovList(proportionalCovList, m_dap.constants.numGrp);
            cvCoef.D = 1;
            cvCoef.detD = 1;
            for k = 1:m_dap.constants.numGrp
                dR(k) = quadraticRule(m_dap.meanList{k}, tempInvCovList{k}, dataElem, m_dap.constants.aprioriProb(k), cvCoef);
            end
        else            
            tempInvCovList = m_dap.invCovList;
            tempInvCovList{i} = inv(cov(m_dap.datasets.dataPerGroups{i}([1:j-1 j+1:end],:)));
            for k = 1:m_dap.constants.numGrp
                
%                cvCoef = quadraticCVCoefficent(m_dap.invCovList{k},dataElem, k, i, m_dap);  % TODO: check Cross Validation (leave-one-out) functionality

                cvCoef = struct('D', 1, 'detD', 1);
                dR(k) = quadraticRule(m_dap.meanList{k}, tempInvCovList{k}, dataElem, m_dap.constants.aprioriProb(k), cvCoef);
            end
        end
        [dummy, group] = max(dR);
        if group ~= i
            malclassificationList{i} = [malclassificationList{i}, j];
        end
    end
end

disp(['Cross-Validation done ', datestr(now)])