[Scip] SCIP-2.1.1 and SCIP-3.0.1 give different optimal values

[Tobias Achterberg]

Yes, this should be a bug somewhere in SCIP.

CPLEX finds the 2205487 solution, this solution looks almost perfectly 
feasible, and the MIP kappa statistics show a 100% stable bases:

MIP - Integer optimal solution:  Objective =  2.2054870000e+06
Solution time =    0.63 sec.  Iterations = 371  Nodes = 0
Deterministic time = 553.33 ticks  (876.25 ticks/sec)

CPLEX> disp sol qual
Incumbent solution:
MILP objective                                 2.2054870000e+06
MILP solution norm |x| (Total, Max)            4.03400e+03  1.00000e+00
MILP solution error (Ax=b) (Total, Max)        0.00000e+00  0.00000e+00
MILP x bound error (Total, Max)                0.00000e+00  0.00000e+00
MILP x integrality error (Total, Max)          9.99201e-16  1.11022e-16
MILP slack bound error (Total, Max)            3.33067e-16  1.11022e-16

Branch-and-cut subproblem optimization:
Max condition number:                    2.6097e+01
Percentage (number) of stable bases:     100.00%   (20)
Percentage (number) of suspicious bases:   0.00%   (0)
Percentage (number) of unstable bases:     0.00%   (0)
Percentage (number) of ill-posed bases:    0.00%   (0)


Regards,

Tobias


On 05/31/13 17:06, Victor Miller wrote:
> This looks like it might be a bug in presolve.  As a guess I set
> maxrounds to 3 in 3.0.1 and then it found the right optimal solution.
>
> Victor
>
>
> On Fri, May 31, 2013 at 10:39 AM, Masahiro Sakai
> <masahiro.sakai at gmail.com <mailto:masahiro.sakai at gmail.com>> wrote:
> Hi,
>
> I solved a minimization MILP problem attached as rand717_l2.wcnf.lp.gz
> using SCIP-2.1.1 and SCIP-3.0.1, but they delivered differenct results:
>
> * SCIP-2.1.1 returned 2205487 as the optimal value, while
> * SCIP-3.0.1 returned 2223207 as the optimal value.
>
> I'm wondering whether it is a bug of SCIP-3.0.1 or just a numerical
> problem.
>
> The problem rand717_l2.wcnf.lp is converted from a weighted partial
> Max-SAT problem rand717_l2.wcnf.gz. And I used SCIP binaries
> downloaded from the SCIP website: scip-2.1.1.linux.x86_64.gnu.opt.spx.zip
> and scip-3.0.1.linux.x86_64.gnu.opt.spx.zip.
>
> Thanks in advance,
>
> Masahiro
>
> ______________________________
>
>     _________________
>     Scip mailing list
>     Scip at zib.de <mailto:Scip at zib.de>
>     http://listserv.zib.de/mailman/listinfo/scip
>
>
>
> On Fri, May 31, 2013 at 10:39 AM, Masahiro Sakai
> <masahiro.sakai at gmail.com <mailto:masahiro.sakai at gmail.com>> wrote:
>
>     Hi,
>
>     I solved a minimization MILP problem attached as rand717_l2.wcnf.lp.gz
>     using SCIP-2.1.1 and SCIP-3.0.1, but they delivered differenct results:
>
>     * SCIP-2.1.1 returned 2205487 as the optimal value, while
>     * SCIP-3.0.1 returned 2223207 as the optimal value.
>
>     I'm wondering whether it is a bug of SCIP-3.0.1 or just a numerical
>     problem.
>
>     The problem rand717_l2.wcnf.lp is converted from a weighted partial
>     Max-SAT problem rand717_l2.wcnf.gz. And I used SCIP binaries
>     downloaded from the SCIP website:
>     scip-2.1.1.linux.x86_64.gnu.opt.spx.zip
>     and scip-3.0.1.linux.x86_64.gnu.opt.spx.zip.
>
>     Thanks in advance,
>
>     Masahiro
>
>     _______________________________________________
>     Scip mailing list
>     Scip at zib.de <mailto:Scip at zib.de>
>     http://listserv.zib.de/mailman/listinfo/scip
>
>
>
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> http://listserv.zib.de/mailman/listinfo/scip