Randomness in IPOPT solve

[madabhushipb]

Hi!

I am working on the problem of optimizing the cost of delivering power to consumers by solving an NLP containing AC - Power Flow equations as constraints (AC-OPF problem). I observed that two successive runs of IPOPT on the same machine take different paths to converge to the optimal solution from the same starting point. In one of the paths, I observed that the linear solver ma27 requests for more memory to be allocated, and consequently takes longer time to solve. Furthermore, I observed that in this case there is a difference in the way IPOPT did regularization internally (see iterations 52 to end in the cases below). Therefore, I am wondering if there is some randomness involved in the way IPOPT chooses to regularize the hessian matrix. I attaching the solver iterations from the two runs here.

Case 1 - [Short Run]

COIN-OR Ipopt    32.2.0 rc62c018 Released Aug 26, 2020 LEG x86 64bit/Linux    

COIN-OR Interior Point Optimizer (Ipopt Library 3.13.2)
written by A. Waechter, commercially supported by GAMS Development Corp.
Number of OpenMP threads: 8.

List of user-set options:

                                    Name   Value                used

******************************************************************************
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 http://projects.coin-or.org/Ipopt
******************************************************************************

This is Ipopt version 3.13.2, running with linear solver ma27.

Number of nonzeros in equality constraint Jacobian...:  2421293
Number of nonzeros in inequality constraint Jacobian.:  1090954
Number of nonzeros in Lagrangian Hessian.............:   447168

Total number of variables............................:   877288
                     variables with only lower bounds:        0
                variables with lower and upper bounds:   168182
                     variables with only upper bounds:        0
Total number of equality constraints.................:   830857
Total number of inequality constraints...............:   567431
        inequality constraints with only lower bounds:   246349
   inequality constraints with lower and upper bounds:        0
        inequality constraints with only upper bounds:   239962

iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
   0  0.0000000e+00 2.64e-02 1.00e+00   0.0 0.00e+00    -  0.00e+00 0.00e+00   0
   1  1.8624434e-03 2.29e-02 8.68e-01  -7.1 2.89e-02  -2.0 1.95e-01 1.36e-01h  1
   2  1.8848881e-03 2.28e-02 2.94e+00  -3.3 2.18e-01    -  8.11e-02 1.26e-03f  1
   3  6.6805437e-03 1.26e-02 2.62e+00  -7.2 5.27e-02    -  9.62e-02 4.12e-01h  1
   4  9.5772285e-03 5.76e-03 2.16e+00  -7.3 6.48e-02    -  2.05e-01 4.33e-01h  1
   5  1.1083311e-02 2.41e-03 2.00e+00  -7.4 9.33e-02    -  1.27e-01 4.05e-01h  1
   6  1.1921994e-02 1.50e-03 1.53e+00  -7.4 6.95e-02    -  2.60e-01 3.80e-01h  1
   7  1.2512857e-02 8.65e-04 1.04e+00  -7.5 5.46e-02    -  2.53e-01 4.23e-01h  1
   8  1.2896347e-02 4.76e-04 7.21e-01  -7.7 4.25e-02    -  2.81e-01 4.53e-01h  1
   9  1.3228937e-02 1.76e-04 1.06e+00  -7.8 3.53e-02    -  1.71e-01 6.35e-01h  1
iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
  10  1.3432625e-02 2.25e-05 1.97e+00  -7.9 2.22e-02    -  2.51e-01 8.79e-01h  1
  11  1.3460561e-02 2.00e-06 1.64e+00  -8.0 4.57e-03    -  4.48e-01 9.13e-01h  1
  12  1.3453243e-02 2.65e-07 6.39e-01  -8.3 1.38e-03    -  6.74e-01 8.69e-01h  1
  13  1.3418418e-02 4.73e-08 5.19e-02  -8.8 4.32e-03    -  9.24e-01 9.62e-01h  1
  14  1.2950756e-02 7.08e-06 3.48e-02  -9.9 5.40e-02    -  3.82e-01 9.97e-01h  1
  15  1.2547667e-02 8.40e-06 5.72e-02 -10.1 6.04e-02    -  3.55e-01 6.61e-01h  1
  16  1.2395295e-02 6.76e-06 2.65e-02 -10.3 1.72e-02    -  4.11e-01 2.53e-01h  1
  17  1.2094165e-02 6.75e-06 9.08e-03 -10.5 2.67e-02    -  3.14e-01 3.43e-01h  1
  18  1.1921564e-02 6.61e-06 1.93e-02 -10.7 4.77e-02    -  1.87e-01 1.52e-01h  1
  19  1.1724130e-02 6.94e-06 2.43e-02 -10.8 8.05e-02    -  1.72e-01 1.51e-01h  1
iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
  20  1.1565466e-02 7.15e-06 1.62e-02 -10.9 1.12e-01    -  9.58e-02 1.11e-01h  1
  21  1.1386973e-02 7.60e-06 2.06e-02 -10.9 1.25e-01    -  1.41e-01 1.22e-01h  1
  22  1.1252138e-02 7.74e-06 2.65e-02 -11.0 1.43e-01    -  1.14e-01 8.98e-02h  1
  23  1.1134368e-02 7.82e-06 2.57e-02 -11.0 1.55e-01    -  8.18e-02 7.75e-02h  1
  24  1.1040214e-02 7.79e-06 5.11e-02 -11.0 1.61e-01    -  1.55e-01 6.26e-02h  1
  25  1.0908076e-02 8.08e-06 6.75e-02 -11.0 1.76e-01    -  1.69e-01 8.52e-02h  1
  26  1.0815024e-02 8.12e-06 8.94e-02 -11.0 1.89e-01    -  1.85e-01 5.98e-02h  1
  27  1.0621096e-02 9.57e-06 8.53e-02 -11.0 2.01e-01    -  1.63e-01 1.23e-01h  1
  28  1.0525424e-02 9.62e-06 1.11e-01 -11.0 1.98e-01    -  2.82e-01 6.49e-02h  1
  29  1.0472134e-02 9.63e-06 1.22e+01  -7.2 4.29e-01    -  7.07e-02 5.17e-02h  1
iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
  30  1.0423093e-02 1.04e-05 1.12e+01  -6.9 4.19e-01    -  1.01e-01 8.96e-02h  1
  31  1.0369760e-02 1.11e-05 1.01e+01  -6.8 5.80e-01    -  1.11e-01 1.02e-01h  1
  32  1.0287157e-02 1.10e-05 9.47e+00  -7.0 3.68e-01    -  1.37e-01 5.85e-02h  1
  33  1.0261472e-02 1.08e-05 1.04e+01  -6.7 5.20e-01    -  1.65e-01 5.28e-02h  1
  34  1.0294998e-02 1.07e-05 1.11e+01  -6.5 5.57e-01    -  1.30e-01 5.51e-02h  1
  35  1.0435227e-02 1.12e-05 1.73e+01  -5.9 1.62e+00    -  1.45e-01 2.38e-02f  1
  36  1.0570948e-02 1.17e-05 2.10e+01  -5.9 1.42e+00    -  1.46e-01 3.64e-02f  1
  37  1.0749552e-02 1.29e-05 2.55e+01  -5.9 1.17e+00    -  2.56e-01 6.52e-02f  1
  38  1.0909357e-02 1.41e-05 3.45e+01  -5.9 1.12e+00    -  6.36e-01 9.28e-02f  1
  39  1.1035325e-02 1.48e-05 3.41e+01  -5.9 9.51e-01    -  7.77e-01 1.29e-01h  1
iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
  40  1.1399528e-02 2.22e-05 2.72e+01  -5.9 5.34e-01    -  1.00e+00 2.97e-01h  1
  41  1.1897885e-02 2.86e-05 1.68e+01  -5.9 3.66e-01    -  1.00e+00 5.29e-01h  1
  42  1.2395900e-02 1.45e-05 7.38e-07  -5.9 1.73e-01    -  1.00e+00 1.00e+00h  1
  43  1.2410554e-02 3.81e-11 2.57e-06  -5.9 7.70e-04  -2.5 1.00e+00 1.00e+00h  1
  44  1.2406302e-02 3.57e-10 3.78e+00  -6.6 7.66e-03  -3.0 1.00e+00 6.00e-03h  8
  45  1.2404537e-02 4.21e-10 5.68e+00  -6.7 2.56e-02  -3.4 1.00e+00 8.09e-04h 11
  46  1.2403455e-02 4.45e-10 6.01e+00  -6.7 4.95e-02  -3.9 1.00e+00 4.46e-04h 12
  47  1.0047554e-02 2.00e-08 6.61e-06  -6.7 7.37e-02  -4.4 1.00e+00 1.00e+00H  1
  48  1.0047083e-02 2.00e-08 9.51e-01  -7.2 6.60e-02  -4.9 7.84e-01 5.32e-04h 11
  49  1.0046738e-02 2.00e-08 2.19e+00  -7.1 1.17e-01  -5.3 1.00e+00 3.42e-04h 12
iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
  50  1.0046299e-02 2.00e-08 2.21e+00  -7.1 1.43e-01  -5.8 1.00e+00 4.35e-04h 12
  51  9.0780494e-03 1.02e-08 4.41e-06  -7.1 1.65e-01  -6.3 1.00e+00 1.00e+00H  1
  52  9.0738103e-03 1.03e-08 4.52e-01  -7.5 9.26e-02  -6.8 9.93e-01 1.06e-02h  7
  53  9.0707890e-03 1.03e-08 4.14e-01  -7.6 1.92e-01  -7.2 1.00e+00 6.39e-03h  8
  54  9.0691634e-03 1.03e-08 2.32e-01  -7.7 2.87e-01  -7.7 1.00e+00 3.24e-03h  9
  55  9.0673747e-03 1.03e-08 1.25e-01  -7.7 3.81e-01  -8.2 1.00e+00 3.39e-03h  9
  56  9.0370150e-03 6.62e-08 7.58e-02  -7.7 4.57e-01  -8.7 1.00e+00 5.68e-02h  5
  57  8.7964842e-03 4.23e-06 1.43e-01  -7.9 3.89e-01  -9.2 1.00e+00 4.26e-01h  2
  58  8.5338278e-03 1.24e-06 4.03e-02  -7.9 3.98e-01  -9.6 1.00e+00 8.91e-01H  1
  59  8.4888630e-03 4.80e-07 3.69e-02  -8.0 2.42e-02 -10.1 1.00e+00 8.89e-01h  1
iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
  60  8.4361048e-03 2.92e-07 2.61e-01  -8.4 6.53e-02 -10.6 1.00e+00 7.99e-01H  1
  61  8.4324563e-03 2.78e-07 1.01e+00  -8.8 8.67e-02 -11.1 1.00e+00 8.19e-02h  4
  62  8.4032836e-03 3.36e-06 6.25e-01  -8.9 1.64e-01 -11.5 1.00e+00 5.67e-01h  1
  63  8.3957070e-03 3.77e-07 8.06e-08  -8.7 3.28e-02 -12.0 1.00e+00 1.00e+00h  1
  64  8.3864297e-03 1.34e-07 3.99e-01  -8.9 5.45e-02 -12.5 1.00e+00 8.39e-01H  1
  65  8.3787339e-03 1.56e-09 4.06e-08  -9.1 1.09e-01 -13.0 1.00e+00 1.00e+00H  1
  66  8.3732855e-03 1.88e-12 8.44e-04  -9.4 1.59e-01 -13.5 1.00e+00 1.00e+00H  1
  67  8.3705742e-03 8.43e-07 7.08e-09  -9.6 1.82e-01 -13.9 1.00e+00 1.00e+00s 22
  68  8.3690431e-03 5.07e-07 3.19e-09  -9.8 1.44e-01 -14.4 1.00e+00 1.00e+00h  1
  69  8.3678527e-03 3.94e-07 2.32e-09 -10.2 1.21e-01 -14.9 1.00e+00 1.00e+00h  1
iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
  70  8.3674143e-03 1.72e-07 5.50e-10 -10.4 1.58e-01 -15.4 1.00e+00 1.00e+00h  1
  71  8.3671848e-03 1.11e-07 2.29e-02 -10.6 1.71e-01 -15.8 9.94e-01 1.00e+00h  1
  72  8.3669932e-03 1.21e-07 1.00e-10 -10.9 1.96e-01 -16.3 1.00e+00 1.00e+00h  1
  73  8.3669423e-03 5.22e-08 1.29e-11 -11.0 1.53e-01 -16.8 1.00e+00 1.00e+00h  1
  74  8.3669424e-03 7.01e-10 8.33e-13 -11.0 2.85e-02 -17.3 1.00e+00 1.00e+00h  1

Number of Iterations....: 74

                                   (scaled)                 (unscaled)
Objective...............:   8.3669423863235622e-03    8.3669423863235622e-03
Dual infeasibility......:   8.3331730506853384e-13    8.3331730506853384e-13
Constraint violation....:   7.0136785268459789e-10    7.0136785268459789e-10
Complementarity.........:   1.0264098760206865e-11    1.0264098760206865e-11
Overall NLP error.......:   7.0136785268459789e-10    7.0136785268459789e-10


Number of objective function evaluations             = 232
Number of objective gradient evaluations             = 75
Number of equality constraint evaluations            = 232
Number of inequality constraint evaluations          = 232
Number of equality constraint Jacobian evaluations   = 75
Number of inequality constraint Jacobian evaluations = 75
Number of Lagrangian Hessian evaluations             = 74
Total CPU secs in IPOPT (w/o function evaluations)   =   1609.116
Total CPU secs in NLP function evaluations           =    115.572

EXIT: Optimal Solution Found.

Case 2 - [Long Run]

COIN-OR Ipopt    32.2.0 rc62c018 Released Aug 26, 2020 LEG x86 64bit/Linux    

COIN-OR Interior Point Optimizer (Ipopt Library 3.13.2)
written by A. Waechter, commercially supported by GAMS Development Corp.
Number of OpenMP threads: 8.

List of user-set options:

                                    Name   Value                used

******************************************************************************
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 http://projects.coin-or.org/Ipopt
******************************************************************************

This is Ipopt version 3.13.2, running with linear solver ma27.

Number of nonzeros in equality constraint Jacobian...:  2421293
Number of nonzeros in inequality constraint Jacobian.:  1090954
Number of nonzeros in Lagrangian Hessian.............:   447168

Total number of variables............................:   877288
                     variables with only lower bounds:        0
                variables with lower and upper bounds:   168182
                     variables with only upper bounds:        0
Total number of equality constraints.................:   830857
Total number of inequality constraints...............:   567431
        inequality constraints with only lower bounds:   246349
   inequality constraints with lower and upper bounds:        0
        inequality constraints with only upper bounds:   239962

iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
   0  0.0000000e+00 2.64e-02 1.00e+00   0.0 0.00e+00    -  0.00e+00 0.00e+00   0
   1  1.8624434e-03 2.29e-02 8.68e-01  -7.1 2.89e-02  -2.0 1.95e-01 1.36e-01h  1
   2  1.8848881e-03 2.28e-02 2.94e+00  -3.3 2.18e-01    -  8.11e-02 1.26e-03f  1
   3  6.6805437e-03 1.26e-02 2.62e+00  -7.2 5.27e-02    -  9.62e-02 4.12e-01h  1
   4  9.5772285e-03 5.76e-03 2.16e+00  -7.3 6.48e-02    -  2.05e-01 4.33e-01h  1
   5  1.1083311e-02 2.41e-03 2.00e+00  -7.4 9.33e-02    -  1.27e-01 4.05e-01h  1
   6  1.1921994e-02 1.50e-03 1.53e+00  -7.4 6.95e-02    -  2.60e-01 3.80e-01h  1
   7  1.2512857e-02 8.65e-04 1.04e+00  -7.5 5.46e-02    -  2.53e-01 4.23e-01h  1
   8  1.2896347e-02 4.76e-04 7.21e-01  -7.7 4.25e-02    -  2.81e-01 4.53e-01h  1
   9  1.3228937e-02 1.76e-04 1.06e+00  -7.8 3.53e-02    -  1.71e-01 6.35e-01h  1
iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
  10  1.3432625e-02 2.25e-05 1.97e+00  -7.9 2.22e-02    -  2.51e-01 8.79e-01h  1
  11  1.3460561e-02 2.00e-06 1.64e+00  -8.0 4.57e-03    -  4.48e-01 9.13e-01h  1
  12  1.3453243e-02 2.65e-07 6.39e-01  -8.3 1.38e-03    -  6.74e-01 8.69e-01h  1
  13  1.3418418e-02 4.73e-08 5.19e-02  -8.8 4.32e-03    -  9.24e-01 9.62e-01h  1
  14  1.2950756e-02 7.08e-06 3.48e-02  -9.9 5.40e-02    -  3.82e-01 9.97e-01h  1
  15  1.2547667e-02 8.40e-06 5.72e-02 -10.1 6.04e-02    -  3.55e-01 6.61e-01h  1
  16  1.2395295e-02 6.76e-06 2.65e-02 -10.3 1.72e-02    -  4.11e-01 2.53e-01h  1
  17  1.2094165e-02 6.75e-06 9.08e-03 -10.5 2.67e-02    -  3.14e-01 3.43e-01h  1
  18  1.1921564e-02 6.61e-06 1.93e-02 -10.7 4.77e-02    -  1.87e-01 1.52e-01h  1
  19  1.1724130e-02 6.94e-06 2.43e-02 -10.8 8.05e-02    -  1.72e-01 1.51e-01h  1
iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
  20  1.1565466e-02 7.15e-06 1.62e-02 -10.9 1.12e-01    -  9.58e-02 1.11e-01h  1
  21  1.1386973e-02 7.60e-06 2.06e-02 -10.9 1.25e-01    -  1.41e-01 1.22e-01h  1
  22  1.1252138e-02 7.74e-06 2.65e-02 -11.0 1.43e-01    -  1.14e-01 8.98e-02h  1
  23  1.1134368e-02 7.82e-06 2.57e-02 -11.0 1.55e-01    -  8.18e-02 7.75e-02h  1
  24  1.1040214e-02 7.79e-06 5.11e-02 -11.0 1.61e-01    -  1.55e-01 6.26e-02h  1
  25  1.0908076e-02 8.08e-06 6.75e-02 -11.0 1.76e-01    -  1.69e-01 8.52e-02h  1
  26  1.0815024e-02 8.12e-06 8.94e-02 -11.0 1.89e-01    -  1.85e-01 5.98e-02h  1
  27  1.0621096e-02 9.57e-06 8.53e-02 -11.0 2.01e-01    -  1.63e-01 1.23e-01h  1
  28  1.0525424e-02 9.62e-06 1.11e-01 -11.0 1.98e-01    -  2.82e-01 6.49e-02h  1
  29  1.0472134e-02 9.63e-06 1.22e+01  -7.2 4.29e-01    -  7.07e-02 5.17e-02h  1
iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
  30  1.0423093e-02 1.04e-05 1.12e+01  -6.9 4.19e-01    -  1.01e-01 8.96e-02h  1
  31  1.0369760e-02 1.11e-05 1.01e+01  -6.8 5.80e-01    -  1.11e-01 1.02e-01h  1
  32  1.0287157e-02 1.10e-05 9.47e+00  -7.0 3.68e-01    -  1.37e-01 5.85e-02h  1
  33  1.0261472e-02 1.08e-05 1.04e+01  -6.7 5.20e-01    -  1.65e-01 5.28e-02h  1
  34  1.0294998e-02 1.07e-05 1.11e+01  -6.5 5.57e-01    -  1.30e-01 5.51e-02h  1
  35  1.0435227e-02 1.12e-05 1.73e+01  -5.9 1.62e+00    -  1.45e-01 2.38e-02f  1
  36  1.0570948e-02 1.17e-05 2.10e+01  -5.9 1.42e+00    -  1.46e-01 3.64e-02f  1
  37  1.0749552e-02 1.29e-05 2.55e+01  -5.9 1.17e+00    -  2.56e-01 6.52e-02f  1
  38  1.0909357e-02 1.41e-05 3.45e+01  -5.9 1.12e+00    -  6.36e-01 9.28e-02f  1
  39  1.1035325e-02 1.48e-05 3.41e+01  -5.9 9.51e-01    -  7.77e-01 1.29e-01h  1
iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
  40  1.1399528e-02 2.22e-05 2.72e+01  -5.9 5.34e-01    -  1.00e+00 2.97e-01h  1
  41  1.1897885e-02 2.86e-05 1.68e+01  -5.9 3.66e-01    -  1.00e+00 5.29e-01h  1
  42  1.2395900e-02 1.45e-05 7.38e-07  -5.9 1.73e-01    -  1.00e+00 1.00e+00h  1
  43  1.2410554e-02 3.81e-11 2.57e-06  -5.9 7.70e-04  -2.5 1.00e+00 1.00e+00h  1
  44  1.2406302e-02 3.57e-10 3.78e+00  -6.6 7.66e-03  -3.0 1.00e+00 6.00e-03h  8
  45  1.2404537e-02 4.21e-10 5.68e+00  -6.7 2.56e-02  -3.4 1.00e+00 8.09e-04h 11
  46  1.2403455e-02 4.45e-10 6.01e+00  -6.7 4.95e-02  -3.9 1.00e+00 4.46e-04h 12
  47  1.0047554e-02 2.00e-08 6.61e-06  -6.7 7.37e-02  -4.4 1.00e+00 1.00e+00H  1
  48  1.0047083e-02 2.00e-08 9.51e-01  -7.2 6.60e-02  -4.9 7.84e-01 5.32e-04h 11
  49  1.0046737e-02 2.00e-08 2.19e+00  -7.1 1.17e-01  -5.3 1.00e+00 3.42e-04h 12
iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
  50  1.0046299e-02 2.00e-08 2.21e+00  -7.1 1.43e-01  -5.8 1.00e+00 4.35e-04h 12
  51  9.0780505e-03 1.02e-08 4.41e-06  -7.1 1.65e-01  -6.3 1.00e+00 1.00e+00H  1
  52  9.0738105e-03 1.03e-08 4.52e-01  -7.5 2.24e-01    -  9.93e-01 1.06e-02h  7
  53  9.0707885e-03 1.03e-08 4.14e-01  -7.6 5.23e-01    -  1.00e+00 6.39e-03h  8
  54  9.0691628e-03 1.03e-08 2.32e-01  -7.7 4.84e-01    -  1.00e+00 3.24e-03h  9
  55  9.0673741e-03 1.03e-08 1.25e-01  -7.7 5.05e-01    -  1.00e+00 3.40e-03h  9
  56  9.0370152e-03 6.62e-08 7.60e-02  -7.7 5.33e-01    -  1.00e+00 5.68e-02h  5
  57  8.7964924e-03 4.23e-06 1.43e-01  -7.9 4.05e-01    -  1.00e+00 4.26e-01h  2
  58  8.5338534e-03 1.24e-06 4.03e-02  -7.9 3.90e-01    -  1.00e+00 8.91e-01H  1
  59  8.4887392e-03 4.63e-07 3.18e-02  -8.0 2.15e-02    -  1.00e+00 9.03e-01h  1
MA27BD returned iflag=-4 and requires more memory.
 Increase liw from 57859270 to 115718540 and la from 122516755 to 247360824 and factorize again.
iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr  ls
  60  8.4363305e-03 2.87e-07 2.60e-01  -8.4 6.50e-02    -  1.00e+00 7.98e-01H  1
MA27BD returned iflag=-4 and requires more memory.
 Increase liw from 115718540 to 231437080 and la from 247360824 to 509162616 and factorize again.
  61  8.4326506e-03 2.73e-07 1.01e+00  -8.8 8.63e-02    -  1.00e+00 8.24e-02h  4
  62  8.4253376e-03 3.70e-07 1.22e+00  -8.8 1.52e-01    -  1.00e+00 1.55e-01h  3
  63  8.4179917e-03 5.22e-07 1.21e+00  -8.9 1.43e-01    -  1.00e+00 1.76e-01h  3
  64  8.4144198e-03 5.28e-07 1.43e+00  -8.9 1.34e-01    -  1.00e+00 9.90e-02h  4
  65  8.3863874e-03 5.14e-07 3.23e-01  -8.9 1.34e-01    -  1.00e+00 8.92e-01H  1
  66  8.3788835e-03 2.29e-09 4.59e-08  -9.1 7.49e-02    -  1.00e+00 1.00e+00H  1
  67  8.3731754e-03 1.35e-09 2.40e-02  -9.4 3.39e-02  -6.8 9.93e-01 1.00e+00H  1
  68  8.3706485e-03 8.39e-10 7.10e-09  -9.6 4.33e-02  -7.2 1.00e+00 1.00e+00H  1

Number of Iterations....: 68

                                   (scaled)                 (unscaled)
Objective...............:   8.3706485040455775e-03    8.3706485040455775e-03
Dual infeasibility......:   7.0959566387022511e-09    7.0959566387022511e-09
Constraint violation....:   8.3890083768523027e-10    8.3890083768523027e-10
Complementarity.........:   5.7995909976889530e-10    5.7995909976889530e-10
Overall NLP error.......:   7.0959566387022511e-09    7.0959566387022511e-09


Number of objective function evaluations             = 216
Number of objective gradient evaluations             = 69
Number of equality constraint evaluations            = 216
Number of inequality constraint evaluations          = 216
Number of equality constraint Jacobian evaluations   = 69
Number of inequality constraint Jacobian evaluations = 69
Number of Lagrangian Hessian evaluations             = 68
Total CPU secs in IPOPT (w/o function evaluations)   =   7102.896
Total CPU secs in NLP function evaluations           =    105.023

EXIT: Optimal Solution Found.

Thanks,
Pranav

[svigerske]

There is no explicit randomness.
Sometimes, using a multithreading feature can introduce nonderterminism.
You could try whether enabling the Strict CNR mode from Intel MKL via an environment variable does help: https://software.intel.com/content/www/us/en/develop/articles/introduction-to-the-conditional-numerical-reproducibility-cnr.html

[damienhocking]

We see similar random behaviour on the same machine and from machine to machine as well, different iteration counts, different path taken etc. I haven't tried CNR though, it might be worth a shot. Damien

…

On 5/13/2021 8:31 AM, Stefan Vigerske wrote: There is no explicit randomness. Sometimes, using a multithreading feature can introduce nonderterminism. You could try whether enabling the Strict CNR mode from Intel MKL via an environment variable does help: https://software.intel.com/content/www/us/en/develop/articles/introduction-to-the-conditional-numerical-reproducibility-cnr.html <https://software.intel.com/content/www/us/en/develop/articles/introduction-to-the-conditional-numerical-reproducibility-cnr.html> — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#469 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAKBTUWHQSNTERRUJ7S45TLTNPPDXANCNFSM442QQ2VA>.

[GTRay]

I have also had a similar issue. The convergence property is different on different machines. Enabling the Strict CNR mode doesn't seem to help. Is there any advice on solving this issue?