Can GlobalSearch has different results, when it runs by stepwise-run and loop-run?

Hi,

Now I try to compare results using stepwise-run GlobalSearch function, loop-run GlobalSearch function and sum-square error minimization (SSE) to find a global min. Because my file needs large data files, I can not show. But I can describe it briefly like pseudocode.

And my environment is R2011a at Win7 32&64bit and I have ML, OPT tbx, Global OPT(GOPT) tbx, Sym math Tbx.

% SSE

>>x = linspace(0,1e2,1e6);

>>for i = 1 : length(x)

mytrialsse(i) = sum((experimentdatavector-modelfunction(x(i))).^2);

>>end

>>[Val Idx] = min(mytrialsse);

>>tau = x(Idx);

% GlobalSearch

>>load experimentaldatavector

>>mytrialsse = sum((experimentdatavector-modelfunction(x)).^2)

>>export mytrialsse to myobjfile1 by matlabFunction cmd with Symbolic Math Tbx

>>x0 = 0; lb = 0; ub = 150;

>>problem1 = createOptimProblem('fmincon', 'objective', @(x)myobjfile1(x),... 'x0',x0,'lb',lb, 'ub', ub);

>>gs = GlobalSearch('Display','final', 'NumTrialPoints', 1e4,'TolFun',1e-9,'TolX',1e-9);

>>[trialtau ~] = run(gs, problem1);

>>delete objective file

In this case, I already know global min as as solution by SSE. So when I compare a solution of stepwise-run(run by line) from GOPT Tbx, those results are similar. But when I make a for loop of GlobalSearch for other experimental data, the results are so different to those of SSE and stepwise-run.

Could anyone advise to me?