myNMIACC.m

function [res]= myNMIACC(U,Y,numclass)

stream = RandStream.getGlobalStream;
reset(stream);
U_normalized = U ./ repmat(sqrt(sum(U.^2, 2)), 1, size(U, 2));
maxIter = 50;
% tmp1 = zeros(maxIter,1);
% tmp2 = zeros(maxIter,1);
% tmp3 = zeros(maxIter,1);
% for iter = 1:maxIter
%     indx = litekmeans(U_normalized,numclass,'MaxIter',100, 'Replicates',1);
%     indx = indx(:);
%     [newIndx] = bestMap(Y,indx);
%     tmp1(iter) = mean(Y==newIndx);
%     tmp2(iter) = MutualInfo(Y,newIndx);
%     tmp3(iter) = purFuc(Y,newIndx);
% end
% res = [max(tmp1);max(tmp2);max(tmp3)];
%U_normalized = gather(U_normalized);
indx = litekmeans(U_normalized,numclass, 'MaxIter',100, 'Replicates',maxIter);
% indx = kmeans(U_normalized,numclass, 'MaxIter',100, 'Replicates',maxIter);
indx = indx(:);
[newIndx] = bestMap(Y,indx);
res(1) = mean(Y==newIndx);
res(2) = MutualInfo(Y,newIndx);
res(3) = purFuc(Y,newIndx);