[Cbc] LP relaxation vs. best possible on root node w/o cuts or preprocessing

[John Forrest]

At least two possibilities.

1)  Some of the LP solves do a quick tightening of bounds - so that 
might change the bound on an integer variable.

2) More likely is that strong branching has found one way on a variable 
to be infeasible.  That message is printed after the first branching 
variable has been decided, but before doing a branch.

John Forrest




On 12/03/12 20:33, Miles Lubin wrote:
> Hello,
>
> I have a model where I observe the following output when I run cbc
> with the options "-presolve off -preprocess off -cuts off -heur off":
>
> Continuous objective value is 1.12693e+06 - 5.63 seconds
> Cbc0010I After 0 nodes, 1 on tree, 1e+50 best solution, best possible
> 1294936.3 (11.40 seconds)
>
> Note that the lower bound at the root node is significantly greater
> than the continuous objective value. What is cbc doing here if both
> integer preprocessing and cuts are disabled?
>
> I can provide an mps file on request.
>
> Thanks in advance,
> Miles Lubin
> _______________________________________________
> Cbc mailing list
> Cbc at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/cbc
>
>