[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Help-glpk] Dynamically change Backtracking technique
From: |
Andrew Makhorin |
Subject: |
Re: [Help-glpk] Dynamically change Backtracking technique |
Date: |
Thu, 11 Dec 2008 21:17:27 +0300 |
> I would like to know if it is possible to dinamically change the
> control parameter which determines the backtracking technique.
> In particular, I would like to spent some time with glpk searching
> with "best local bound" optionĀ (to improve dual bound) and afterwards
> change to "depth first search", to intensity the search for feasible
> solutions.
> Is it possible ? What is the best way of doing this ?
You need to pass your own callback routine to the mip solver
routine glp_intopt. Being called (for reason GLP_IHEUR, for example)
the callback routine may check optimal solution to the current LP
relaxation with routines glp_ios_get_prob and glp_get_obj_val, and
depending on that, change the parameter bt_tech in the same parameter
block (glp_iocp), which was passed to glp_intopt. For more details
please see the glpk reference manual, chapter "Branch-and-Cut API
Routines".
(It seems to me that using the depth first search first would be more
appropriate.)