> ILOG OPL Studio allows to set integer values of a feasible solution inside > the mathematical model (like we could do in GMPL) and use an optimization > profile parameter to force the solver to branch according to these values to > find the first feasible solution. Afterwards a normal branching heuristic is > used. This can be a way to start with a good upper bound.
> This might be a feature worth considering for implementation in GLPSOL. Yes. Ampl allows specifying an initial estimation of variables in the same way as computable values for parameters, e.g. var x, >= 0, <= 10, := 3.14159; Such data are useless for the simplex solver, however, it can be passed to the mip solver (to specify some integer feasible solution) or to the interior-point solver (to specify a point to start the search from). So it is reasonable to implement this feature in Mathprog. For mip models there might be additional attributes of variables and constraints, for example, branching priority, etc. _______________________________________________ Help-glpk mailing list [email protected] http://lists.gnu.org/mailman/listinfo/help-glpk
