[Clp] Help: incorrect solution from CBC

[John Forrest]

Greg,

The problems seem to stem from the range of numbers in model e.g. a pair 
of constraints -

  -1E+15 C190 +1 C225 <= +0

and

  -4000000 C190 +1 C225 >= +0

where C190 is binary

I can see this causing problems (for both codes) - if C190 is 1.0e-9 is 
this integer feasible?

Solving two ways I get

ma.sol:Optimal - objective value 1664550445.89107466
mb.sol:Optimal - objective value 1676264519.23880672

while changing e15 to e7 (or e6 - which will fix C190 to an exact zero) 
I get

mc.sol:Optimal - objective value 1647162379.14471364

If you want you can send me the glpsol solution and I can see if I can 
duplicate that solution.

John Forrest




On 27/08/14 02:32, Greg Gruber wrote:
> Hello,
>
> I am attempting to use CBC v2.8.9, and it is reporting that it has 
> found the optimal solution, but the objective value that it returns is 
> about 20% higher than the result I get when using another solver 
> (glpsol).  I did not set any options, just used the defaults:
>
>   "cbc model.lp"
>
> When running glpsol, the only option that I used was enabling cuts:
>
>   "glpsol --lp model.lp --cuts"
>
> I have attached the .lp file with the model that I am trying to solve. 
> I appreciate any help.
>
> Thanks,
> Greg
>
>
>
>
> _______________________________________________
> Clp mailing list
> Clp at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/clp

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/clp/attachments/20140827/7268ba67/attachment.html>