Hello Andreas,the MIP solver rejects branches of the tree of all solutions in the course of the solution process. Hence rerunning the solver is not avoidable.
If you use the GLPK library you could add the last solution as additional constraints to the problem and resolve the problem.
If you use the glpsol executable, you will have to restart it and will need some external storage. Using an ODBC database is an alternative to files. I think it is preferable because you can use it to accumulate the solutions, while in the case of files, you have to write all old solutions to the file again.
In the appendix the shiftcover model is changed to only use binary variables. This makes excluding possible solutions much easier.
The idea for the conversion is replacing integer variables (crew[s]) by sums of powers of two times binary variables (sum{i in I} 2**i * crew[s, i]).
Best regards Xypron Andreas F wrote:
From: [email protected] On Wed, 2011-02-16 at 14:48 +0000, Andreas F wrote:However, is it possible to generate multiple, different solutions [..] ?Currently glpk does not have such a feature. However, you may obtain an alternate mip solution by introducing a packing inequality to disable the current optimal solution. See http://lists.gnu.org/archive/html/help-glpk/2010-02/msg00033.html for example.The solution is indeed elegant, but it requires use of files and one has to re-run the solver. Is it possible to generate more than one solution without using files? Best regards, Andreas
shiftcover2.gz
Description: GNU Zip compressed data
_______________________________________________ Help-glpk mailing list [email protected] http://lists.gnu.org/mailman/listinfo/help-glpk
