[Couenne] Couenne Result

[Pietro Belotti]

Hi Priscilla,

it seems that only the output is wrong. The values you see are printed
by Cbc, which is the branch-and-bound code that Couenne uses.
Sometimes Couenne and Cbc "disagree" on whether a solution is feasible
or not, so I would guess (it happens often) that Couenne finds the
solution with value 38291.9 but Cbc refuses to set it as the best
solution and therefore does not print it. Couenne saves that solution
and uses it to solve the problem faster. If you use the option
"output_level 2" in the option file couenne.opt you should see more
statistics at the end that will show the right value.

The minus sign here is probably due to the problem being a
maximization one: Couenne takes by default minimization problems.

Pietro

On Sat, Oct 18, 2014 at 9:49 PM, Priscilla Lusie
<priscilla.lusie at gmail.com> wrote:
> Hi,
> I'm having a problem running Couenne for a non linear problem. It is
> printing the 'best possible' value greater than the 'best objective' but
> comparing to the result of others solvers, the 'best objective' is wrong.
> The correct value is one of the 'best possible' value found.
> The optimal value is 38291,70938 (found with another solvers). When I run
> display Profit (objective function), Couenne prints the correct value
> (38291). What's wrong?
> The couenne result is:
> Cbc0010I After 5600 nodes, 195 on tree, -26661.508 best solution, best
> possible -39139.591 (86.86 seconds)
> Cbc0010I After 5700 nodes, 184 on tree, -26661.508 best solution, best
> possible -38988.877 (86.96 seconds)
> Cbc0010I After 5800 nodes, 165 on tree, -26661.508 best solution, best
> possible -38773.25 (87.05 seconds)
> Cbc0010I After 5900 nodes, 137 on tree, -26661.508 best solution, best
> possible -38501.2 (87.15 seconds)
> Cbc0010I After 6000 nodes, 89 on tree, -26661.508 best solution, best
> possible -38094.493 (87.21 seconds)
> Cbc0010I After 6100 nodes, 39 on tree, -26661.508 best solution, best
> possible -34536.558 (87.25 seconds)
> Cbc0001I Search completed - best objective -26661.50836015483, took 57960
> iterations and 6178 nodes (87.27 seconds)
> Cbc0035I Maximum depth 86, 0 variables fixed on reduced cost
>         "Finished"
>
> couenne: Optimal
>
> ampl: display Profit;
> Profit = 38291.9
>
>
> Thank's,
> Priscilla Lusie
>
> _______________________________________________
> Couenne mailing list
> Couenne at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/couenne