[Couenne] couenne optimal result is inferior than my heuristic?

[Pietro Belotti]

Hi Renan,

the inferior objective function might be due to a sufficient gap
reached by the BB that triggers termination, though I believe the gap
target for Couenne should be zero. Also, by how much is the solution
violated? If you need a lower tolerance, please set the parameter
feas_tolerance in the option file. Unfortunately I cannot help more as
I would need the instance in order to reproduce the problem. Please
post a .nl file, which you can produce by writing "write gmodel;" in
AMPL after loading the model.

Regards,
Pietro

On Wed, Jul 15, 2015 at 9:34 AM, renan maglasang
<maglasangrenan at gmail.com> wrote:
> Hello to you Pietro, and to all members.
>
> Any help would mean a lot. I itemized my questions so you can address it
> separately.
> I copied the couenne result below for your reference.
>
> QUESTIONS:
>
> Question 1.
> -the heuristic i used can find a better result which is
>
> facings:
> 1-4
> 2-3
> 3-1
> 4-1
> 5-15
> 6-9
> profit = 1807.64
>
> -(although the profit difference is slight but the bigger the better and the
> variable values are different too)
> -i checked capacity capacity of both solution, its okay both are 164 which
> are less than max of 166
>
> Question 2.
> -i found some solution by couenne which violates the constraints too, it
> says optimal but when i try to manually compute, capacity constraint is
> violated.
>
> Question 3.
> This warning below...does this have something to do with all these
> discrepancies of couenne results?
> Error in an AMPL evaluation. Run with "halt_on_ampl_error yes" to see
> details.
>
>
>
> _____________________________________________________________________________________________________
>
> ampl: model irion.txt;
>
> ampl: data bulb.txt;
>
> ampl: option solver couenne;
>
> ampl: solve;
>
> couenne:
>
> ANALYSIS TEST: Problem size before reformulation: 12 variables (12 integer),
> 1 constraints.
>
> Problem size after  reformulation: 109 variables (19 integer), 1
> constraints.
>
>
>
> ******************************************************************************
>
> This program contains Ipopt, a library for large-scale nonlinear
> optimization.
>
>  Ipopt is released as open source code under the Common Public License
> (CPL).
>
>          For more information visit http://projects.coin-or.org/Ipopt
>
> ******************************************************************************
>
>
>
> NOTE: You are using Ipopt by default with the MUMPS linear solver.
>
>       Other linear solvers might be more efficient (see Ipopt
> documentation).
>
>
>
>
>
> Error in an AMPL evaluation. Run with "halt_on_ampl_error yes" to see
> details.
>
> Cbc0012I Integer solution of -1807.27 found by Init Rounding NLP after 0
> iterations and 0 nodes (0.00 seconds)
>
> Cbc0031I 30 added rows had average density of 2
>
> Cbc0013I At root node, 30 cuts changed objective from -9.99e+012 to
> -9.99e+012 in 13 passes
>
> Cbc0014I Cut generator 0 (Couenne convexifier cuts) - 327 row cuts average
> 2.0 elements, 16 column cuts (46 active)
>
> Cbc0010I After 0 nodes, 1 on tree, -1807.27 best solution, best possible
> -9.99e+012 (0.09 seconds)
>
> Cbc0010I After 100 nodes, 19 on tree, -1807.27 best solution, best possible
> -9.99e+012 (1.25 seconds)
>
> Cbc0010I After 200 nodes, 28 on tree, -1807.27 best solution, best possible
> -9.99e+012 (1.50 seconds)
>
> ……………………….(I deleted some parts to make it shorter)
>
> Cbc0010I After 58600 nodes, 163 on tree, -1807.27 best solution, best
> possible -1834.51 (90.14 seconds)
>
> Cbc0010I After 58700 nodes, 127 on tree, -1807.27 best solution, best
> possible -1832.28 (90.25 seconds)
>
> Cbc0010I After 58800 nodes, 92 on tree, -1807.27 best solution, best
> possible -1829.64 (90.37 seconds)
>
> Cbc0010I After 58900 nodes, 51 on tree, -1807.27 best solution, best
> possible -1825.32 (90.48 seconds)
>
> Cbc0010I After 59000 nodes, 14 on tree, -1807.27 best solution, best
> possible -1819.16 (90.62 seconds)
>
> Cbc0001I Search completed - best objective -1807.268742209032, took 520060
> iterations and 59028 nodes (90.64 seconds)
>
> Cbc0035I Maximum depth 121, 1253 variables fixed on reduced cost
>
>  Couenne convexifier cuts was tried 94303 times and created 1520652 cuts of
> which 138882 were active after adding rounds
>
>                 "Finished"
>
>
>
> couenne: Optimal
>
> ampl: display facings;
>
> facings [*] :=
>
> 1   4
>
> 2   2
>
> 3   1
>
> 4   1
>
> 5  15
>
> 6  10
>
> ;
>
> ampl: display profit;
>
> profit = 1807.27
>
>
> _______________________________________________
> Couenne mailing list
> Couenne at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/couenne