finalize_solution() Returns Wrong or Incomplete Multiplier and Constraint Values When Terminating Due to Exceeding the Maximum Time

[yanhaoy]

When using the C++ interface, we assume that most optimization information can be obtained from the function finalize_solution(). However, I have noticed that when the optimization terminates due to exceeding the maximum time (max_wall_time or max_cpu_time), the returned multiplier and constraint values are wrong or incomplete in finalize_solution(). For example, solve the hs071 problem and set max_wall_time=0.001. the output is:

******************************************************************************
This program contains Ipopt, a library for large-scale nonlinear optimization.
 Ipopt is released as open source code under the Eclipse Public License (EPL).
         For more information visit https://github.com/coin-or/Ipopt
******************************************************************************

This is Ipopt version 3.14.5, running with linear solver ma27.

Number of nonzeros in equality constraint Jacobian...:        4
Number of nonzeros in inequality constraint Jacobian.:        4
Number of nonzeros in Lagrangian Hessian.............:       10

Total number of variables............................:        4
                     variables with only lower bounds:        0
                variables with lower and upper bounds:        4
                     variables with only upper bounds:        0
Total number of equality constraints.................:        1
Total number of inequality constraints...............:        1
        inequality constraints with only lower bounds:        1
   inequality constraints with lower and upper bounds:        0
        inequality constraints with only upper bounds:        0

iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
   0  1.6109693e+01 1.12e+01 5.28e-01   0.0 0.00e+00    -  0.00e+00 0.00e+00   0
   1  1.7410406e+01 7.49e-01 2.25e+01  -0.3 7.97e-01    -  3.19e-01 1.00e+00f  1
   2  1.8001613e+01 7.52e-03 4.96e+00  -0.3 5.60e-02   2.0 9.97e-01 1.00e+00h  1

Number of Iterations....: 2

                                   (scaled)                 (unscaled)
Objective...............:   1.8001612675074764e+01    1.8001612675074764e+01
Dual infeasibility......:   4.9625297838268363e+00    4.9625297838268363e+00
Constraint violation....:   7.5231790452647829e-03    7.5231790452647829e-03
Variable bound violation:   0.0000000000000000e+00    0.0000000000000000e+00
Complementarity.........:   5.3326217059024505e-01    5.3326217059024505e-01
Overall NLP error.......:   4.9625297838268363e+00    4.9625297838268363e+00


Number of objective function evaluations             = 3
Number of objective gradient evaluations             = 3
Number of equality constraint evaluations            = 3
Number of inequality constraint evaluations          = 3
Number of equality constraint Jacobian evaluations   = 3
Number of inequality constraint Jacobian evaluations = 3
Number of Lagrangian Hessian evaluations             = 2
Total seconds in IPOPT                               = 0.001

EXIT: Maximum wallclock time exceeded.


Solution of the primal variables, x
x[0] = 1.17223
x[1] = 4.33696
x[2] = 4.28941
x[3] = 1.1938


Solution of the bound multipliers, z_L and z_U
z_L[0] = 0
z_L[1] = 0
z_L[2] = 0
z_L[3] = 0
z_U[0] = 0
z_U[1] = 0
z_U[2] = 0
z_U[3] = 0


Objective value
f(x*) = 0

Final value of the constraints:
g(0) = 0
g(1) = 40


*** The problem FAILED!

It stops at iteration 2 with incomplete multiplier and constraint values. Conversely, if we set the termination condition as max_iter=2, the output becomes:

******************************************************************************
This program contains Ipopt, a library for large-scale nonlinear optimization.
 Ipopt is released as open source code under the Eclipse Public License (EPL).
         For more information visit https://github.com/coin-or/Ipopt
******************************************************************************

This is Ipopt version 3.14.5, running with linear solver ma27.

Number of nonzeros in equality constraint Jacobian...:        4
Number of nonzeros in inequality constraint Jacobian.:        4
Number of nonzeros in Lagrangian Hessian.............:       10

Total number of variables............................:        4
                     variables with only lower bounds:        0
                variables with lower and upper bounds:        4
                     variables with only upper bounds:        0
Total number of equality constraints.................:        1
Total number of inequality constraints...............:        1
        inequality constraints with only lower bounds:        1
   inequality constraints with lower and upper bounds:        0
        inequality constraints with only upper bounds:        0

iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
   0  1.6109693e+01 1.12e+01 5.28e-01   0.0 0.00e+00    -  0.00e+00 0.00e+00   0
   1  1.7410406e+01 7.49e-01 2.25e+01  -0.3 7.97e-01    -  3.19e-01 1.00e+00f  1
   2  1.8001613e+01 7.52e-03 4.96e+00  -0.3 5.60e-02   2.0 9.97e-01 1.00e+00h  1

Number of Iterations....: 2

                                   (scaled)                 (unscaled)
Objective...............:   1.8001612675074764e+01    1.8001612675074764e+01
Dual infeasibility......:   4.9625297838268363e+00    4.9625297838268363e+00
Constraint violation....:   7.5231790452647829e-03    7.5231790452647829e-03
Variable bound violation:   0.0000000000000000e+00    0.0000000000000000e+00
Complementarity.........:   5.3326217059024505e-01    5.3326217059024505e-01
Overall NLP error.......:   4.9625297838268363e+00    4.9625297838268363e+00


Number of objective function evaluations             = 3
Number of objective gradient evaluations             = 3
Number of equality constraint evaluations            = 3
Number of inequality constraint evaluations          = 3
Number of equality constraint Jacobian evaluations   = 3
Number of inequality constraint Jacobian evaluations = 3
Number of Lagrangian Hessian evaluations             = 2
Total seconds in IPOPT                               = 0.001

EXIT: Maximum Number of Iterations Exceeded.


Solution of the primal variables, x
x[0] = 1.17223
x[1] = 4.33696
x[2] = 4.28941
x[3] = 1.1938


Solution of the bound multipliers, z_L and z_U
z_L[0] = 0.835262
z_L[1] = 0.158199
z_L[2] = 0.162115
z_L[3] = 0.129356
z_U[0] = 0.135017
z_U[1] = 0.774348
z_U[2] = 0.749755
z_U[3] = 0.137127


Objective value
f(x*) = 18.0016

Final value of the constraints:
g(0) = 26.0331
g(1) = 40.0075


*** The problem FAILED!

It reproduces the same decision variable values and optimization process, but now we get the full multiplier and constraint values.

Any ideas on how to fix this, or is it deliberately designed this way? Thanks!

[svigerske]

Thank you for the detailed report.

I think when the termination due to the CPU timelimit was added, it was forgotten to add that the full solution should be passed on to finalize_solution(). And when the termination due to the wallclock timelimit was added by mostly copy-pasting the other timelimit, this was forgotten, too.

It turns out that this is easy to fix: c967728.