For the time being, my branch-and-bound algorithm is not optimal. Anyway, I won't need an optimal solution (it may be difficult to reach it due to the size of the problems I am facing). So even if my branch-and-bound algorithm was correct, as I have to stop the run before the optimal solution, I may get different final solutions on Windows and on Linux. But fortunately, this does not occur very often.
2009/6/23 Andrew Makhorin <[email protected]> > > > If your branch-and-price algorithm is correct, it must not depend > on which *optimal* basic solution is reported by lp solver. Don't you > expect a particular optimal basis, do you? > >
_______________________________________________ Help-glpk mailing list [email protected] http://lists.gnu.org/mailman/listinfo/help-glpk
