Though PROC OPTLSO does conduct a global search of the feasible region using a GA and LHS, it does not currently attempt to calculate a global optimality certificate such as one might find in a branch and bound algorithm.  Typically the computation of such certificates either require much stronger assumptions on the objective and constraint functions and/or significantly longer running time.  So unfortunately there is no way to guarantee the returned point is globally optimal.  However, there are many things we can try to increase the probability that a global optimal point is found such as increasing popsize, nlocal, and/or nglobal, as you suggest. In general I have found that increasing popsize tends to be preferrable, but as you pointed out it is not guaranteed to improve the objective.   If you use seed=0, a random seed will be generated each time.  One possibility is to run with (or without) seed=0 and perform a small grid search over a set of popsizes (something similar could be added to search on nglobal/nlocal).  For example:   %let popsize=200; %macro loopSolve(ntimes); proc delete data=pall; %do i=1 %to &ntimes.; %let popsize = %eval(&popsize. + 10*&i.); proc optlso /* your other options here */ popsize=&popsize. seed=0 primalout=pout; performance nthreads=2; run; data pout; set pout; _sol_=&i.; run; proc append base=pall data=pout; run; %end; %mend; %loopSolve(10); proc transpose data=pall out=pallrow (drop=_label_ _name_); by _sol_; var _value_; id _id_; run; proc print data=pallrow; run; I've attached a documentation example using the above code.  Hopefully this help and please let us know if you have more questions.   Best,   --Josh   ... View more

Hi Davi,   PROC GA provides a hook for an update routine, called each iteration.  You can use this to do things like report the iteration count and current best objective.  Below is an example the prints the best objective each iteration to Output.   Best,   --Josh   OPTIONS NODATE NONUMBER; TITLE "Iteration history"; TITLE "Iteration history output"; proc ga seed = 555; call SetEncoding('R3R2'); npoints = 3; array cvxhull[3,2] /nosym ( -2 0 0 2 2 -2 ); /* Objective function */ function sixhump(selected[*],cvxhull[*,*],npoints); /* Function has global minimum value of -1.0316 * at x = {-0.0898 0.7126} and * x = { 0.0898 -0.7126} */ array w[1] /nosym; call dynamic_array(w,npoints); array x[2] /nosym; call ReadMember(selected,1,w); /* make sure that weights add up to 1 */ sum = 0; do i = 1 to npoints; sum + w[i]; end; /* if all weights 0, then reinitialize */ if sum=0 then do; sum = npoints; do i = 1 to npoints; w[i] = 1; end; end; /* re-normalize weights */ do i = 1 to npoints; w[i] = w[i] / sum; end; call WriteMember(selected,1,w); /* convert weights to x-coordinate form */ x[1] = 0; x[2] = 0; do i = 1 to npoints; x[1] + w[i] * cvxhull[i,1]; x[2] + w[i] * cvxhull[i,2]; end; /* write out x coordinates to second segment */ call WriteMember(selected,2,x); /* compute objective value */ r = (4 - 2.1*x[1]**2 + x[1]**4/3)*x[1]**2 + x[1]*x[2] + (-4 + 4*x[2]**2)*x[2]**2; return(r); endsub; /*************************************/ subroutine gethistory(iteration, maxiter, popsize, iterArray[*], bestArray[*]); outargs iteration, iterArray, bestArray; iteration = iteration+1; array objValues[1] /nosym; /* dynamically allocate array to fit populationSize */ call dynamic_array(objValues, popsize); /* read in current objective values */ call GetObjValues(objValues, popsize); /* find best value */ current_best = objValues[1]; do i = 2 to popsize; /* for a minimization problem, use < here */ if(objValues[i] < current_best) then current_best = objValues[i]; end; *put iteration current_best; iterArray[iteration]=iteration; bestArray[iteration]=current_best; if (iteration = maxiter) then do; put +1; do i = 1 to maxiter; put "iter=" iterArray[i] ", bestObj=" bestArray[i] +1; end; end; endsub; /************************************/ call SetObjFunc('sixhump',0); array lower[1] /nosym; array upper[1] /nosym; call dynamic_array(lower, npoints); call dynamic_array(upper, npoints); do i = 1 to npoints; lower[i] = 0; upper[i] = 1; end; call SetBounds(lower, upper, 1); array delta[3] /nosym (0.01 0.01 0.01); call SetMut('delta', 'nchange', 1, 'delta', delta); call SetMutProb(0.05); call SetCross('Twopoint', 'alpha', 0.9); call SetCrossProb(0.8); call SetSel('tournament', 'size', 2); call SetElite(3); iteration=0; popsize=20; maxiter=40; array iterArray[1] /nosym; array bestArray[1] /nosym; call dynamic_array(iterArray, maxiter+1); call dynamic_array(bestArray, maxiter+1); call SetUpdateRoutine('gethistory'); call Initialize('DEFAULT', popsize); call ContinueFor(maxiter); run;   ... View more