[Bonmin] unexpected unfeasible problem

[Nicolas BERGER]

Thank you Pierre !! Now it works indeed... when Ampl uses Bonmin. No idea
about the problem with MINOS ?

And, by the way, do you know other free MINLP solvers ?


Nicolas


> Hi Nicolas,
>
> The problem is that Ipopt was not able to solve the root node relaxation.
> Apparently some of the options changed by bonmin are the cause for this
> (although I don't really understand why).
> You can try with the option file that I attach (put the file in the
> directory where you run ampl). With it I get a solution of same objective
> value as Baron.
> Best regards,
> Pierre
> On 9/10/07, Nicolas BERGER <Nicolas.Berger at univ-nantes.fr> wrote:
>>
>> Hi
>>
>>
>> I am trying to solve this problem :
>>
>> ------------------------------------------------------------------
>> var x0 integer >= 1 <= 10 ;
>> var y0 integer >= 1 <= 10 ;
>> var x1 >= 1 <= 10 ;
>> var y1 >= 1 <= 10 ;
>> var x2 >= 1 <= 10 ;
>> var y2 >= 1 <= 10 ;
>>
>> maximize dist:   (x0-x1)**2 + (y0-y1)**2 + (x0-x2)**2 +
>>                  (y0-y2)**2 + (x1-x2)**2 + (y1-y2)**2 ;
>>
>>
>> subject to
>> e01: (x0-x1)**2 + (y0-y1)**2 >= 12.5 ;
>> e10: (x0-x1)**2 + (y0-y1)**2 <= 64 ;
>> e02: (x0-x2)**2 + (y0-y2)**2 >= 12.5 ;
>> e20: (x0-x2)**2 + (y0-y2)**2 <= 64 ;
>> e12: (x1-x2)**2 + (y1-y2)**2 >= 12.5 ;
>> e21: (x1-x2)**2 + (y1-y2)**2 <= 64 ;
>>
>> solve ;
>> display x0,y0,x1, y1 ,x2, y2  ;
>> ------------------------------------------------------------------------
>>
>>
>>
>> Here is the solution to this small problem (given by gams/baron) :
>>
>>
>> -----------------------------------------------------------------------------------
>>                             LOWER          LEVEL          UPPER
>> MARGINAL
>>
>> ---- VAR objvar            -INF          192.0000        +INF
>> .
>> ---- VAR x0                 1.0000         1.0000        10.0000
>> EPS
>> ---- VAR y0                 1.0000         6.0000        10.0000
>> EPS
>> ---- VAR x1                 1.0000         7.9282        10.0000
>>           .
>> ---- VAR y1                 1.0000        10.0000        10.0000
>>           .
>> ---- VAR x2                 1.0000         7.9282        10.0000
>>           .
>> ---- VAR y2                 1.0000         2.0000        10.0000
>>           .
>>
>> -----------------------------------------------------------------------------------
>>
>>
>>
>> Now here is what ampl returns when I call it on this problem :
>>
>> --------------------------------------------------------
>> MINOS 5.5: ignoring integrality of 2 variables
>> MINOS 5.5: infeasible problem (or bad starting guess).
>> 0 iterations
>> Nonlin evals: obj = 7, grad = 6, constrs = 7, Jac = 6.
>> x0 = 1
>> y0 = 1
>> x1 = 1
>> y1 = 1
>> x2 = 1
>> y2 = 1
>> --------------------------------------------------------
>>
>>
>>
>> And this is what I get when I add "option solver bonmin" in the problem
>> file :
>>
>>
>> -------------------------------------------------------------------------------------
>> bonmin:
>>
>>
>> ******************************************************************************
>> 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
>>
>> ******************************************************************************
>>
>> IpOp0008I
>>               Num      Status      Obj             It       time
>> IpOp0009I     1        FAILED      -0             3000     2.81218
>> IpOp0010I     r1       FAILED      -0             3000     2.78817
>> Ipopt exited with error code -1 Maximum Iterations Exceeded
>> can't open /tmp/at23241.sol
>>
>> -------------------------------------------------------------------------------------
>>
>> What am I doing wrong ? Where to start debugging ? :)
>>
>>
>>
>>
>> >From Nantes, France,
>> BERGER Nicolas
>>
>>
>>
>> _______________________________________________
>> Bonmin mailing list
>> Bonmin at list.coin-or.org
>> http://list.coin-or.org/mailman/listinfo/bonmin
>>
>