[ADMB Users] NLS example Really bad starting values

[dave fournier]

There is so much noise about good and bad starting values for this example
that I thought I would try ADMB with really bad starting values, say 
(1000,1000,1000,1000)

It easily converges to the correct value. So it appears all the noise 
about starting values
was really about bad software.

It would be nice to work up the other examples. Maybe it will do itself.

dave at shithead ~ $ ./nlstest -iprint 200 -crit 1.e-10 -maxfn 2000
cmpdiff.tmp

Initial statistics: 4 variables; iteration 0; function evaluation 0; phase 1
Function value   1.5414908e+06; maximum gradient component mag 3.0839e+03
Var   Value    Gradient   |Var   Value    Gradient   |Var   Value Gradient
   1 1000.000  3.08391e+03 |  2 1000.000  3.07639e+03 |  3 1000.000 
-1.96666e+03
   4 1000.000 -1.11169e+03 |

Intermediate statistics: 4 variables; iteration 200; function evaluation 
261; phase 1
Function value   1.7632404e-03; maximum gradient component mag -1.6493e-01
Var   Value    Gradient   |Var   Value    Gradient   |Var   Value Gradient
   1  0.02115 -1.64932e-01 |  2 3171.925 -1.10090e-06 |  3 287.7796 
7.73631e-06
   4 121.3562  1.07607e-05 |
4 variables; iteration 400; function evaluation 529; phase 1
Function value   9.4393223e-04; maximum gradient component mag -2.2130e-02
Var   Value    Gradient   |Var   Value    Gradient   |Var   Value Gradient
   1  0.00164 -2.21304e-02 |  2 1465.606 -2.42582e-08 |  3  7.37513 
2.51702e-06
   4  4.97857  2.43657e-06 |
4 variables; iteration 600; function evaluation 789; phase 1
Function value   9.4194729e-04; maximum gradient component mag -3.8763e-03
Var   Value    Gradient   |Var   Value    Gradient   |Var   Value Gradient
   1  0.00614 -3.87629e-03 |  2 381.7429 -5.48863e-08 |  3  7.17742 
1.60209e-06
   4  4.85359  1.55681e-06 |
4 variables; iteration 800; function evaluation 1048; phase 1
Function value   8.9388331e-04; maximum gradient component mag -9.9181e-03
Var   Value    Gradient   |Var   Value    Gradient   |Var   Value Gradient
   1  0.07519 -9.91808e-03 |  2 18.40742 -3.42472e-05 |  3  4.12940 
7.05283e-05
   4  2.93613  9.89697e-05 |
   ic > imax  in fminim is answer attained ?
Var   Value    Gradient   |Var   Value    Gradient   |Var   Value Gradient
   1  0.19281  7.12727e-10 |  2  0.19128 -1.60142e-10 |  3  0.12306 
4.12968e-11
   4  0.13606  3.75012e-10 |
Function minimizer not making progress ... is minimum attained?
Minimprove criterion =   0.0000e+00

  - final statistics:
4 variables; iteration 850; function evaluation 1142
Function value   3.0751e-04; maximum gradient component mag 7.1273e-10
Exit code = 1;  converg criter   1.0000e-10
Var   Value    Gradient   |Var   Value    Gradient   |Var   Value Gradient
   1  0.19281  7.12727e-10 |  2  0.19128 -1.60142e-10 |  3  0.12306 
4.12968e-11
   4  0.13606  3.75012e-10 |
Estimating row 1 out of 4 for hessian
Estimating row 2 out of 4 for hessian
Estimating row 3 out of 4 for hessian
Estimating row 4 out of 4 for hessian