[Coin-discuss] Feasibility of linear relaxation

[Laszlo Ladanyi]

Hi Mathieu,

What you describe suspiciously looks like a tolerance problem.
I will look into it early next week, and will give a more detailed explanation
then, too.

Thanks for the bug report,
--Laci

On Thu, 18 May 2006, mathieu lacroix wrote:

> Hi,
> I use BCP and CLP as an branch and cut solver for a mixed integer linear 
> problem. I have tested my problem for many data. Allmost work well. But 
> there is an error for one data file. I have an error : "LP : no var/cut 
> in pool but couldn't select branching object.". The problem seems to 
> come from the lp solver. Indeed, test_feasibility method is called after 
> the LP is solved. Bcp has found that lp program has been solved to 
> optimality (getLpProblemPointer()->lp_solver->isProvenOptimal () is 
> equal to one) and an lp solution is send to method test_feasibility. But 
> the given solution is not feasible (and the lp optimal value is negative 
> although variables and cost are positive). In order to understand, I 
> have change in the file BCP_lp_main_loop.cpp the command  
> "p.lp_solver->resolve();" (used to solve the linear relaxation) by the 
> command "p.lp_solver->initialSolve();". And now, it works fine (I obtain 
> optimal solution). Is it normal that the two methods don't give the same 
> thing? Does anybody have the same problem as me? if yes, how to correct 
> it? I can give more informations if necessary.
> Thank you
> 
> Best regards,
> 
> Mathieu LACROIX
> 
> -- 
> Mathieu LACROIX
> mathieu.lacroix at isima.fr
> Université Blaise Pascal - Clermont Fd II
> Laboratoire LIMOS
> Bâtiment ISIMA - Bureau D112
> Complexe scientifique des Cézeaux
> 63 177 Aubière, Cedex - France
> Tel : 04.73.40.79.48
> Fax : 04.73.40.76.39
> 
> _______________________________________________
> Coin-discuss mailing list
> Coin-discuss at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/coin-discuss
>