function [y DCy cpu] = NNKlite(P,T,PT,NFS,K)

if nargin < 5
    K = 5;
end

[N na] = size(P);
n = size(PT,1);
[q, Q, cs, class] = nc(T);

% N = number of elements in training set
% n = number of elements in testing set
% na = number of attributes (=features)
% q = number of classes
% Q = classes vector

if nargin < 4
    for i = 1:q
        NFS{i} = 1:na;
    end
end


c(q) = 0; Y(n) = 0; DCY(n) = 0;
t = cputime;
for i = 1:n % Do this for every element to be classified
    X = PT(i,:); % Test element
    d = []; % Reset distance matrix
    for j = 1:q
        d = [d; sum((rm(X(NFS{j}),cs(j),1) - P(class{j},:)).^2,2) rm(j,cs(j),1)];
        % This is the distance matrix, formed for each class j
    end
    ds = ksr(d,K); % k smallest rows of distance matrix
    
    for j = 1:q
        c(j) = sum(ds(:,2)==j); % Class counter
    end
    
    [M ind] = max(c);
    DCY(i) = M/(sum(c));
    Y(i) = Q(ind);
end
cpu = cputime - t;
y = Y; DCy = DCY;