Re: [Help-glpk] Callbacks for column generation

2009-12-09 Thread Andrew Makhorin
> Thanks, another question, can you get the final branch and bound tree > after the mip is solved?, i could add columns in each node and then > add to the initial relaxation and re-optimize. The glpk mip solver does not keep fathomed nodes, so if the mip is solved (to optimality), the final tree i

Re: [Help-glpk] Callbacks for column generation

2009-12-09 Thread Javier Zambrano
Thanks, another question, can you get the final branch and bound tree after the mip is solved?, i could add columns in each node and then add to the initial relaxation and re-optimize. Regards Javier 2009/12/9 Andrew Makhorin > What would happen if i add cols in a callback when the reason for

Re: [Help-glpk] Callbacks for column generation

2009-12-09 Thread Andrew Makhorin
> What would happen if i add cols in a callback when the reason for > calling the calback is GLP_IROWGEN, it would start adding columns to > the global problem?. To represent subproblems, i.e. nodes of the search tree, the glpk mip solver uses the original problem object (glp_prob) passed to it fr

Re: [Help-glpk] Callbacks for column generation

2009-12-09 Thread Andrew Makhorin
> What would happen if i add cols in a callback when the reason for > calling the calback is GLP_IROWGEN, it would start adding columns to > the global problem?. Currently glpk mip solver does not allow adding columns to the current subproblem, i.e. using column generation; glp_add_cols checks tha

[Help-glpk] Callbacks for column generation

2009-12-09 Thread Javier Zambrano
I would like to solve a problem with integer variables and column generation (it #39;s not branch and price in the strict meaning) . The idea is for a heuristic for a monthly problem when the weekly problem can be solve with an integer program, the monthly problem linear relaxation can be solve