> Hello.. I have on my model many agents that need to frequently solve > a (relatively small) MIP problem.
> This problem change slightly from one solve() call to to the other, > so I am wandering if I can pass the previous solved quantities as > initial values. > I am currently running my simulations creating and destroying every > time the glpk problems, but I am wandering if it is best to create > them only the first time and destroy them at the end of the > simulation. If I don't destroy the glpk problems, is glpk smart > enought to take advantage of previously solved problems, even if in > the maintime the structure of the matrix is slighty changed (eg. I > added a row and/or a column) ?? This would be reasonable only in the case of lp problems. In case of mip this would give no gain. _______________________________________________ Help-glpk mailing list [email protected] http://lists.gnu.org/mailman/listinfo/help-glpk
