[Coin-ipopt] No optimal variable values in IPOPT+CUTEr output

Carl Damon Laird claird at andrew.cmu.edu
Fri Nov 11 19:40:22 EST 2005


Hello,

Judging by the path information, I will assume that you are using the 
Fortran IPOPT version. I am not as familiar with the Fortran version as 
the new C++ version, but I will look into the CUTEr interface.

For now, you can increase the print level for IPOPT. The option in the 
Fortran version is "IPRINT". Have a look at the README.IPOPT in the doc 
directory. You could also output this information to a file.


Hope this helps,

Carl.

On Fri, 11 Nov 2005, Lihong Zhang wrote:

> Hi All,
>
> I  have a question on the output of IPOPT if using the CUTEr
> interface (ie, SIF as the input model file). From the output,
> I only can read the optimal objective function value, while I can't
> find the optimal variable values from the output list. As a matter
> of fact, both the optimal objective function value and the
> optimal variable values are output when I run LANCELOT
> (also using SIF as the input file). Besides, IPOPT + AMPL
> also can output both optimal values. Below I give the output
> results from IPOPT+CUTEr and LANCELOT, respectively,
> if I input the same HS65.SIF.
>
>
> Results from IPOPT+CUTEr:
> ------------------------------------------------------------------------------
> [lihong at frosty]tmp> sdipopt --blas none HS65.SIF
>
> Problem name: HS65 
> Double precision version will be formed.
>
> The objective function uses        3 nonlinear groups
>
> There  is        1 nonlinear inequality constraint
>
> There are        3 variables bounded from below and above
>
> ld: warning: symbol `evals_' has differing sizes:
>       (file 
> /homes/lihong/cvs/cuter/CUTEr.large.sun.sol.g77/double/bin/ipoptma
> .o value=0x8; file /homes/lihong/cvs/COIN/Ipopt/lib/libipopt.a(ipopt.o) 
> value=0x
> c);
>       /homes/lihong/cvs/COIN/Ipopt/lib/libipopt.a(ipopt.o) definition taken
> ******************************************************************************
> This program contains IPOPT, a program for large-scale nonlinear 
> optimization.
>  IPOPT is released as open source under the Common Public License (CPL).
>              For more information visit www.coin-or.org/Ipopt
> ******************************************************************************
>
> Number of variables           :        4
>  of which are fixed         :        0
> Number of constraints         :        1
> Number of lower bounds        :        4
> Number of upper bounds        :        3
> Number of nonzeros in Jacobian:        4
> Number of nonzeros in Hessian :        4
>
> ITER     ERR       MU      ||C||    ||D||   ALFA(X) #LS        F         Regu
>   0 .200E+02d .100E+00 .830E+01 .000E+00 .000E+00   0 0.11549921E+03 
> .000E+00
>   1 .999E+01d .100E+00 .752E+01 .666E+00 .930E-01h  1 0.11693020E+03 
> .100E+03
>   2 .100E+02d .100E+00 .746E+01 .256E+01 .759E-02h  1 0.11698136E+03 
> .333E+02
>   3 .208E+03d .100E+00 .506E+01 .778E+01 .100E+01f  1 0.99509892E+02 
> .111E+02
>   4 .733E+02d .100E+00 .202E+01 .196E+01 .100E+01h  1 0.10934095E+03 
> .296E+02
>   5 .743E+02d .100E+00 .132E+01 .115E+01 .100E+01h  1 0.12019884E+03 
> .790E+02
>   6 .202E+02d .100E+00 .653E+00 .886E+00 .100E+01f  1 0.11130700E+03 
> .000E+00
>   7 .658E+02p .100E+00 .658E+02 .563E+02 .999E+00f  1 0.57224133E+01 
> .000E+00
>   8 .995E+01p .100E+00 .995E+01 .668E+02 .932E+00f  1 0.10383396E+01 
> .000E+00
>   9 .154E+01p .100E+00 .154E+01 .135E+01 .100E+01h  1 0.16139967E+01 
> .000E+00
>
> ITER     ERR       MU      ||C||    ||D||   ALFA(X) #LS        F         Regu
>  10 .284E+00p .100E+00 .284E+00 .129E+01 .100E+01h  1 0.10375737E+01 
> .000E+00
>  11 .352E-01p .200E-01 .352E-01 .132E+01 .894E+00h  1 0.95303357E+00 
> .000E+00
>  12 .269E-02c .283E-02 .168E-02 .476E-01 .100E+01h  1 0.95638608E+00 
> .000E+00
>  13 .612E-04c .150E-03 .203E-04 .348E-01 .100E+01h  1 0.95366453E+00 
> .000E+00
>  14 .880E-07c .184E-05 .256E-07 .166E-02 .100E+01h  1 0.95353066E+00 
> .000E+00
>  15 .688E-11c .251E-08 .269E-11 .221E-04 .100E+01h  1 0.95352886E+00 
> .000E+00
>
> Number of iterations taken .............                     15
> Final value of objective function is.... 0.9535288576748209E+00
>
> Errors at final point                      (scaled)       (unscaled)
> Final maximal constraint violation is... 0.268651E-11    0.268651E-11
> Final value for dual infeasibility is... 0.574804E-12    0.574804E-12
> Final value of complementarity error is. 0.251278E-08    0.251278E-08
>
> The objective function was evaluated     16 times.
> The constraints were evaluated           16 times.
>
> EXIT: OPTIMAL SOLUTION FOUND
>
> CPU seconds spent in IPOPT and function evaluations =          0.0000
>
> ************************ CUTEr statistics ************************
> Code used               :  IPOPT
> Problem                 :  HS65     # variables             =               3
> # constraints           =               1
> # objective functions   =        0.3300000E+02
> # objective gradients   =        0.1700000E+02
> # objective Hessians    =        0.1600000E+02
> # Hessian-vector prdct  =        0.0000000E+00
> # constraints functions =        0.3400000E+02
> # constraints gradients =        0.1700000E+02
> # constraints Hessians  =        0.1600000E+02
> Exit code               =               0
> Final f                 =   0.9535289E+00
> Set up time             =            0.00 seconds
> Solve time              =            0.03 seconds
> ******************************************************************
>
>
> Results from LANCELOT:
> ------------------------------------------------------------------------------
> [lihong at frosty]sampleproblems> sdlan HS65
>
> Problem name: HS65 
> Double precision version will be formed.
>
> The objective function uses        3 nonlinear groups
>
> There  is        1 nonlinear inequality constraint
>
> There are        3 variables bounded from below and above
> There  is        1 slack variable
>
> objective function value =   9.53529015445393E-01
>
>           X1                3.65046164957023E+00
>           X2                3.65046164897452E+00
>           X3                4.62041746528368E+00
>           C1                0.00000000000000E+00
>
>
>
> Form the above, we may find the output objective
> function values from both IPOPT and LANCELOT
> converge. However, I hope I can also obtain the
> optimal variable values from IPOPT. I guess it is
> not a big deal. I maybe missed some switches. Does
> anybody have ideas and give me any hints? thanks.
>
> Best regards,
>
> Lihong
>
>



More information about the Coin-ipopt mailing list