> I was wondering whether there's any way to stop the simplex solver from > looking once a certain precision for the objective value (a float) has > been reached... > For instance, say we want to minimize a number between 0 and 1 and the > optimal value is 0.5131231. Is there a way to tell glpsol to stop > optimizing if the improvement would be less than, say, 0.001 ? In this > case, 0.514 would be OK. > I tried to do this by declaring the objective as an integer between 0 and > 1000 and dividing it by 1000.0, but that transforms the problem to MIP and > triggers the MIP solver and I still want it to be solved by simplex > (--nomip doesn't work for this either, since it takes way too long to > solve).
Probably you mean the mip solver, not the simplex one. If you are using glpsol, you can specify the option '--mipgap GAP', where GAP is a relative mip gap tolerance. In your case GAP might be, for example, 0.10 (that corresponds to 10%). If you are using glpk api, see the reference manual for description of parameters passed to glp_intopt. _______________________________________________ Help-glpk mailing list [email protected] http://lists.gnu.org/mailman/listinfo/help-glpk
