Why function value increases in fmincon?

Question

0 个投票

I am estimating parameter values that minimizes the objective function. But iteration results show that function values of iteration counts sometimes increases. Below are the output of each stage. As you see, iteration 0 has the lowest f(x) and after that it kind of oscilating.

Question 1: Is it normal? Or should I kill it? (It is still running!) Question 2: I let matlab print function value each time it enters the function. Even before the iteration 0, I saw a function value (0.0068810247341) smaller than the value reported in Iter 0 below (6.881026e-03). Why does this happen?

                                            First-order      Norm of
 Iter F-count            f(x)  Feasibility   optimality         step
     6    6.881026e-03    0.000e+00    9.440e-02
    12    1.388225e-01    0.000e+00    9.763e-01    3.423e-01
    18    3.454902e-02    0.000e+00    1.529e-01    2.966e-01
    24    2.840788e-02    0.000e+00    1.804e-01    4.293e-01
    30    2.293784e-02    0.000e+00    1.141e-01    1.078e+00
    36    1.884519e-02    0.000e+00    1.156e-01    2.076e+00
    42    2.036587e-02    0.000e+00    1.235e-01    1.908e+00
    48    3.710619e-02    0.000e+00    1.308e-01    3.064e+00
    55    8.061270e-02    0.000e+00    1.469e-01    7.555e+00
    62    9.708605e-02    0.000e+00    1.967e-01    1.184e+01
    69    5.904960e-02    0.000e+00    1.232e-01    1.158e+01
    75    3.024435e-02    0.000e+00    9.845e-02    5.270e+00
    81    2.043228e-02    0.000e+00    2.947e-02    1.795e+00
    87    1.728052e-02    0.000e+00    6.405e-03    1.288e+00
    93    1.697164e-02    0.000e+00    2.939e-03    4.742e-01
    99    1.696332e-02    0.000e+00    2.750e-03    5.989e-02