On 4 February 2013 19:16, Heinrich Schuchardt <[email protected]> wrote: >> Except when I tried it didn't work, because the reason GLP_IROWGEN is >> called with the solution to the LP relaxation, at which point you have >> non-integer variables and determining problematic loops isn't >> possible. > > See the book indicated below for different constraints to add when you see a > non-integer solution.
So I guess the answer to my question "will the solver present integral solutions to me in a callback prior to accepting them" is no. Not what I hoped, but ok. Seems like a useful feature, to have a callback in step 5 of the branch and bound algorithm as listed in the documentation. So after GLP_IROWGEN (where the LP relaxation is solved) but before GLP_IBINGO (where the integer solution is accepted), such that you can still modify the problem. > Literature: > The Traveling Salesman Problem: A Computational Study > Princeton Series in Applied Mathematics > David L. Applegate, Robert E. Bixby, Vasek Chvátal and William J. Cook > ISBN-10: 0691129932 I'll see if I can find these books you suggest. It's of course a question how applicable they'll be given my problem only vaguely resembles the TSP, but they may have useful ideas. Have a nice day, -- Martijn van Oosterhout <[email protected]> http://svana.org/kleptog/ _______________________________________________ Help-glpk mailing list [email protected] https://lists.gnu.org/mailman/listinfo/help-glpk
