[Coin-bcpdiscuss] Possible problem in BCP_lp_user::reduced_cost_fixing
Laszlo Ladanyi
ladanyi at us.ibm.com
Tue Nov 29 14:50:46 EST 2005
On Tue, 29 Nov 2005, Laszlo Ladanyi wrote:
> 2) Because of roundoff problems it should be tested that new_ub is not below
> the current lb. I wonder if this was your problem? You did say that you got
> infeasibility, but it's not clear whether the new_ub only went below the
> current value making the current solution infeasible, or below the current lb
> making the LP relaxation infeasible.
Actually, I just realized that this test is not necessary. The code is
writte in a way that new_ub can never go below the current lb, no matter
what roundoff errors there are.
--Laci
More information about the Coin-bcpdiscuss
mailing list