[Coin-discuss] Negative values in LP solution

[Laszlo Ladanyi]

Hi Viet,

It is normal. To reoptimize the dual simplex algo is used, so it is dual feas,
but not primal feas. If during the simplex algo the obj function goes above
the current best IP solution value then there's no point in finishing the
simplex as the optimal LP obj value will be too high, too. So the simplex
stops, and stops with a primal infeas solution. Which means it might have
negative values.

Hope this helps,
--Laci

On Mon, 19 Sep 2005, Hung Nguyen wrote:

> Hi all,
> I am working with COIN/BCP and CPLEX to solve an 0/1 problem.
> Although all my variables are non-negative (as I precise in BB.cpp),
> at some iterations, LP displays the optimal solution with negative values
> in several variables and usually, after this COIN/BCP performs the 
> fathoming
> operation. I think this is abnormal, but I can not explain myself.
> Another problem is the program sometimes crashs in OsiRowCut->setRow 
> routine although
> the cut to be added seems to have no error. I use a version of COIN 
> prior to June 15, from that time
> the routine has major changes, should I do some update for the routine?
> Thankyou for your help.
> Best regards,
> Viet
> 
> 
> _______________________________________________
> Coin-discuss mailing list
> Coin-discuss at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/coin-discuss
>