Started within Anpass for Typ=4
Settings : Gradient limit=0.10u, maximum iterations=80, with clampings
0: pure eval, 1: goal function call, 2: goal function gradient, 3: Hessian
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.

Real Hessian Matrix H at start : 
 0.281049,  0.000000, -0.297327, -2.176687
 0.000000,  1.000000,  0.000000,  0.000000
-0.297327,  0.000000,  0.921487, -0.662228
-2.176687,  0.000000, -0.662228,  5.758395

RMS value of diagonal matrix entries : 2.96
RMS value of non-diagonal entries : 0.94
Ratio of RMS values : 3.162:1

Real inverse Hessian Matrix iH at start : 
-1.115397, -0.000000, -0.722615, -0.504725
 0.000000,  1.000000,  0.000000,  0.000000
-0.722615,  0.000000,  0.714821, -0.190944
-0.504725, -0.000000, -0.190944, -0.039087

Eigenvalues of real damped start iH :  1.000000000,  1.000000000,  1.000000000,  1.000000000
Eigenvalues of real iH at start :  1.000000000,  0.965979414,  0.152262625, -1.557904418
Eigenvalues of real H at start :  6.567599876,  1.035218748,  1.000000000, -0.641887903

Start values : f=2.576, |g|=2.93
X :  0.000000000,  0.000000000,  0.000000000,  0.000000000
Initial step size t=0.125
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 1 (7 calls) :
Current goal function and gradient : f=1.658, reduction df=0.92, |g|=2.93
X :  0.277307809,  0.000000000, -0.192160119, -0.547668358
G : -1.264035203,  0.000000000,  0.875911703,  2.496403136
S :  1.264035203,  0.000000000, -0.875911703, -2.496403136
Angle between -G and S is 0.00.
Angle between old and new gradient is 0.00.
Current parameter steps : |dx|=0.64, max(dxi)=0.55 for number i=4, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=0.22, clampings=0
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH :  1.000000000,  1.000000000,  1.000000000,  1.000000000
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 2 (14 calls) :
Current goal function and gradient : f=1.416, reduction df=0.24, |g|=0.53
X :  0.246020706,  0.000000000, -1.192160119, -0.590401356
G : -0.072327200,  0.000000000,  0.491027403,  0.183873891
S : -0.013863176,  0.000000000, -0.488064258, -0.018934801
Angle between -G and S is 5.74.
Angle between old and new gradient is 33.14.
Current parameter steps : |dx|=1.00, max(dxi)=1.00 for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=2.26, clampings=1
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH :  1.032809709,  1.000000000,  1.000000000,  0.239759811
Warning : Parabel interpolation shows that start stepsize before GST is quite bad.
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 3 (25 calls) :
Current goal function and gradient : f=1.210, reduction df=0.21, |g|=0.46
X :  0.826763068,  0.000000000, -2.192160119, -0.075664501
G : -0.140947996,  0.000000000,  0.106140589, -0.424425073
S :  0.281144440,  0.000000000, -0.689138535,  0.249190371
Angle between -G and S is 35.48.
Angle between old and new gradient is 95.82.
Current parameter steps : |dx|=1.27, max(dxi)=1.00 for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=4, step size is t=2.07, clampings=6
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH :  3.934321929,  1.000000000,  0.794131923,  0.216205605
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 4 (32 calls) :
Current goal function and gradient : f=0.771, reduction df=0.44, |g|=7.16
X :  0.689519314,  0.000000000, -1.935970192, -0.196467834
G :  2.069377126,  0.000000000, -0.125798707,  6.852668543
S : -0.542420201,  0.000000000,  1.012523985, -0.477443719
Angle between -G and S is 44.31.
Angle between old and new gradient is 177.89.
Current parameter steps : |dx|=0.31, max(dxi)=0.26 for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=1, step size is t=0.25, clampings=1
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH :  3.544503627,  1.000000000,  0.445942863,  0.074351986
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 5 (38 calls) :
Current goal function and gradient : f=0.260, reduction df=0.51, |g|=1.62
X :  1.402866777,  0.000000000, -1.919655205, -0.586148463
G : -1.150912911,  0.000000000,  0.086042513,  1.142356859
S :  0.437161181,  0.000000000,  0.009998324, -0.238808229
Angle between -G and S is 3.78.
Angle between old and new gradient is 47.92.
Current parameter steps : |dx|=0.81, max(dxi)=0.71 for number i=1, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=1.63, clampings=0
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.

Current real Hessian Matrix H  : 
46.339675,  0.000000, -1.582495, 38.829301
 0.000000,  1.000000,  0.000000,  0.000000
-1.582495,  0.000000, -0.091464, -1.496368
38.829301,  0.000000, -1.496368, 41.065446

RMS value of diagonal matrix entries : 30.96
RMS value of non-diagonal entries : 15.88
Ratio of RMS values : 1.950:1

Current real inverse Hessian Matrix : 
 0.101862,  0.000000, -0.116946, -0.100577
 0.000000,  1.000000,  0.000000,  0.000000
-0.116946, -0.000000, -6.715524, -0.134126
-0.100577,  0.000000, -0.134126,  0.114564

Current inverse Hessian Matrix from BFGS : 
 0.274273,  0.000000,  0.216010, -0.122626
 0.000000,  1.000000,  0.000000,  0.000000
 0.216010,  0.000000,  3.340714, -0.042748
-0.122626,  0.000000, -0.042748,  0.088724

Sign mismatches in matrices : 3 of 16 (18.75%)
Normalized average entry difference in matrices : 140.42%
Hint : Angle between S via iH and real iH is 0.46. Length ratio is 5.95:1.
Current eigenvalues of real iH :  1.000000000,  0.209036827,  0.012095026, -6.720229335
Current eigenvalues of BFGS iH :  3.356659360,  1.000000000,  0.320379056,  0.026671986
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 6 (44 calls) :
Current goal function and gradient : f=80.790m, reduction df=0.18, |g|=5.29
X :  1.234277553,  0.000000000, -1.965916028, -0.534657693
G :  3.943971792,  0.000000000, -0.151899764,  3.518231368
S : -0.334566458,  0.000000000, -0.091804918,  0.102183782
Angle between -G and S is 45.47.
Angle between old and new gradient is 95.59.
Current parameter steps : |dx|=0.18, max(dxi)=0.17 for number i=1, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=0.50, clampings=0
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH :  3.347585062,  1.000000000,  0.297859213,  0.021949683
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 7 (50 calls) :
Current goal function and gradient : f=61.042m, reduction df=19.75m, |g|=0.34
X :  1.326473250,  0.000000000, -1.953654349, -0.622829236
G : -0.230488409,  0.000000000,  0.008551457,  0.253336687
S :  0.049971051,  0.000000000,  0.006645961, -0.047789917
Angle between -G and S is 0.87.
Angle between old and new gradient is 90.95.
Current parameter steps : |dx|=0.13, max(dxi)=92.20m for number i=1, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=1.84, clampings=0
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH :  3.339605920,  1.000000000,  0.201535519,  0.014788195
Hint : Start stepsize t good already before linear search.
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 8 (55 calls) :
Current goal function and gradient : f=60.753m, reduction df=0.29m, |g|=0.12
X :  1.317058027,  0.000000000, -1.954298783, -0.616022715
G :  0.106464955,  0.000000000, -0.004439892,  0.059887033
S : -0.009415223,  0.000000000, -0.000644435,  0.006806521
Angle between -G and S is 52.54.
Angle between old and new gradient is 110.21.
Current parameter steps : |dx|=11.64m, max(dxi)=9.42m for number i=1, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=1.00, clampings=0
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH :  3.342402592,  1.000000000,  0.337662609,  0.014750859
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 9 (61 calls) :
Current goal function and gradient : f=60.747m, reduction df=5.75u, |g|=17.75m
X :  1.317943519,  0.000000000, -1.952986249, -0.617497550
G :  0.009835121,  0.000000000, -0.000618839,  0.014764809
S :  0.000524646,  0.000000000,  0.000777665, -0.000873827
Angle between -G and S is 57.48.
Angle between old and new gradient is 9.78.
Current parameter steps : |dx|=2.16m, max(dxi)=1.47m for number i=4, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=1.69, clampings=0
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH :  3.339321060,  1.000000000,  0.300691702,  0.016350641
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 10 (67 calls) :
Current goal function and gradient : f=60.747m, reduction df=0.31u, |g|=2.33m
X :  1.317680084,  0.000000000, -1.952145232, -0.617081679
G : -0.000924002,  0.000000000, -0.000148381, -0.002129482
S : -0.000147662,  0.000000000,  0.000471414,  0.000233107
Angle between -G and S is 59.55.
Angle between old and new gradient is 178.14.
Current parameter steps : |dx|=0.97m, max(dxi)=0.84m for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=1.78, clampings=0
Hint : Switched to double-sided gradient because : |g| is smaller 100dxss=2.00m.
Hint : Double-sided gradient formula used

Current real Hessian Matrix H  : 
25.966096,  0.000004, -1.074220, 21.917625
 0.000004, -0.000002, -0.000002, -0.000002
-1.074220, -0.000002,  0.043547, -0.925222
21.917625, -0.000002, -0.925222, 23.772077

RMS value of diagonal matrix entries : 17.60
RMS value of non-diagonal entries : 8.97
Ratio of RMS values : 1.963:1

Current real inverse Hessian Matrix : 
-1.369205, 37.683702, -40.167579, -0.300947
37.683702, -489420.000665, 967.706681,  2.877528
-40.167579, 967.706681, -1045.736991, -3.666460
-0.300947,  2.877528, -3.666460,  0.176836

Current inverse Hessian Matrix from BFGS : 
 0.180718,  0.000000,  0.202475, -0.161866
 0.000000,  1.000000,  0.000000,  0.000000
 0.202475,  0.000000,  3.487828, -0.109511
-0.161866,  0.000000, -0.109511,  0.187332

Sign mismatches in matrices : 5 of 16 (31.25%)
Warning : Matrix quite bad, although already many iterations.
Normalized average entry difference in matrices : 50.32%
Hint : Angle between S via iH and real iH is 2.68. Length ratio is 0.00:1.
Current eigenvalues of real iH :  0.341798859,  0.021341538, -1045.372079927, -489421.921085050
Current eigenvalues of BFGS iH :  3.504468822,  1.000000000,  0.330677343,  0.020733023

Iteration 11 (78 calls) :
Current goal function and gradient : f=60.746m, reduction df=0.89u, |g|=1.00m
X :  1.318399697,  0.000000000, -1.945705879, -0.617646872
G :  0.000185785,  0.000000000, -0.000213817,  0.000963976
S :  0.000090335,  0.000000000,  0.000808347, -0.000070950
Angle between -G and S is 65.38.
Angle between old and new gradient is 174.44.
Current parameter steps : |dx|=6.50m, max(dxi)=6.44m for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=7.97, clampings=0
Hint : Switched to double-sided gradient because : |g| is smaller 100dxss=2.00m.
Hint : Double-sided gradient formula used
Current eigenvalues of BFGS iH :  5.637060280,  1.000000000,  0.290937774,  0.021345593
Warning : Parabel interpolation shows that start stepsize before GST is quite bad.

Iteration 12 (98 calls) :
Current goal function and gradient : f=60.705m, reduction df=41.24u, |g|=2.71m
X :  1.336326973,  0.000000000, -1.517816138, -0.616985399
G : -0.000448357,  0.000000000, -0.000182702, -0.002667494
S :  0.000334302,  0.000000000,  0.007979140,  0.000012335
Angle between -G and S is 83.18.
Angle between old and new gradient is 176.42.
Current parameter steps : |dx|=0.43, max(dxi)=0.43 for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=5, step size is t=53.63, clampings=0
Hint : Switched to double-sided gradient because : |g| is smaller 100dxss=2.00m.
Hint : Double-sided gradient formula used
Current eigenvalues of BFGS iH : 62.539278888,  1.000000000,  0.217048958,  0.020447025

Iteration 13 (111 calls) :
Current goal function and gradient : f=60.650m, reduction df=54.57u, |g|=0.13
X :  1.332906677,  0.000000000, -1.599913690, -0.616916276
G : -0.096244911,  0.000000000,  0.005158991, -0.090417939
S : -0.012349052,  0.000000000, -0.296414966,  0.000249571
Angle between -G and S is 89.17.
Warning : Angle already quite extreme.
Angle between old and new gradient is 18.81.
Current parameter steps : |dx|=82.17m, max(dxi)=82.10m for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=3, step size is t=0.28, clampings=1
Warning : Switching to single-side gradient although double-sided used in previous interations.
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH : 618.678002257,  1.000000000,  0.328263952,  0.012636652
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 14 (124 calls) :
Current goal function and gradient : f=60.284m, reduction df=0.37m, |g|=73.50m
X :  1.360435288,  0.000000000, -1.054960651, -0.606142029
G : -0.052665604,  0.000000000,  0.002369901, -0.051214607
S :  0.004843532,  0.000000000,  0.095881962,  0.001895679
Angle between -G and S is 88.41.
Warning : Angle already quite extreme.
Angle between old and new gradient is 0.02.
Current parameter steps : |dx|=0.55, max(dxi)=0.54 for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=5, step size is t=5.68, clampings=1
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH : 191.330270529,  1.000000000,  0.328967761,  0.013749142
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 15 (131 calls) :
Current goal function and gradient : f=59.925m, reduction df=0.36m, |g|=0.25
X :  1.358224148,  0.000000000, -1.132525489, -0.606081261
G : -0.200459886,  0.000000000,  0.015473923, -0.145561168
S : -0.004279712,  0.000000000, -0.150128517,  0.000117617
Angle between -G and S is 86.42.
Warning : Angle already quite extreme.
Angle between old and new gradient is 0.96.
Current parameter steps : |dx|=77.60m, max(dxi)=77.56m for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=1, step size is t=0.52, clampings=0
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.

Current real Hessian Matrix H  : 
29.610368,  0.000000, -2.648960, 23.696550
 0.000000,  1.000000,  0.000000,  0.000000
-2.648960,  0.000000,  0.234161, -2.206473
23.696550,  0.000000, -2.206473, 24.200890

RMS value of diagonal matrix entries : 19.13
RMS value of non-diagonal entries : 9.78
Ratio of RMS values : 1.957:1

Current real inverse Hessian Matrix : 
-1.212096,  0.000000, -17.947264, -0.449474
 0.000000,  1.000000,  0.000000,  0.000000
-17.947264, -0.000000, -235.428956, -3.891570
-0.449474, -0.000000, -3.891570,  0.126621

Current inverse Hessian Matrix from BFGS : 
 0.957832,  0.000000, 13.617275,  0.099106
 0.000000,  1.000000,  0.000000,  0.000000
13.617275,  0.000000, 224.725511,  4.105075
 0.099106,  0.000000,  4.105075,  0.300715

Sign mismatches in matrices : 8 of 16 (50.00%)
Warning : Matrix quite bad, although already many iterations.
Normalized average entry difference in matrices : 2168.13%
Hint : Angle between S via iH and real iH is 3.14. Length ratio is 0.70:1.
Current eigenvalues of real iH :  1.000000000,  0.326829405,  0.019611712, -236.860871718
Current eigenvalues of BFGS iH : 225.625870055,  1.000000000,  0.335763872,  0.022424629
Warning : Parabel interpolation shows that start stepsize before GST is quite bad.
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 16 (142 calls) :
Current goal function and gradient : f=58.997m, reduction df=0.93m, |g|=54.93m
X :  1.404400122,  0.000000000, -0.566064406, -0.597382101
G : -0.051251733,  0.000000000,  0.001359582, -0.019712323
S :  0.029023417,  0.000000000,  0.356043087,  0.005467765
Angle between -G and S is 84.90.
Angle between old and new gradient is 3.10.
Current parameter steps : |dx|=0.57, max(dxi)=0.57 for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=4, step size is t=1.59, clampings=1
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH : 134.641908363,  1.000000000,  0.340212477,  0.018995581
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 17 (148 calls) :
Current goal function and gradient : f=57.943m, reduction df=1.05m, |g|=0.56
X :  1.397730546,  0.000000000, -0.702429095, -0.601603584
G : -0.441878174,  0.000000000,  0.050393011, -0.347451459
S : -0.012656387,  0.000000000, -0.258769673, -0.008010811
Angle between -G and S is 87.13.
Warning : Angle already quite extreme.
Angle between old and new gradient is 4.17.
Current parameter steps : |dx|=0.14, max(dxi)=0.14 for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=0.53, clampings=0
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH : 82.938061924,  1.000000000,  0.378452200,  0.011322064
Warning : Parabel interpolation shows that start stepsize before GST is quite bad.
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 18 (160 calls) :
Current goal function and gradient : f=53.417m, reduction df=4.53m, |g|=0.20
X :  1.507565271,  0.000000000,  0.187369426, -0.588023923
G : -0.151103573,  0.000000000,  0.010959568, -0.128777515
S :  0.034517724,  0.000000000,  0.279636697,  0.004267676
Angle between -G and S is 85.66.
Warning : Angle already quite extreme.
Angle between old and new gradient is 0.12.
Current parameter steps : |dx|=0.90, max(dxi)=0.89 for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=4, step size is t=3.18, clampings=2
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH : 36.741156114,  1.000000000,  0.364515521,  0.013666765
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 19 (166 calls) :
Current goal function and gradient : f=47.696m, reduction df=5.72m, |g|=0.93
X :  1.472866048,  0.000000000, -0.054939816, -0.576580717
G : -0.603159978,  0.000000000,  0.103904949, -0.705808454
S : -0.042918725,  0.000000000, -0.299707106,  0.014153857
Angle between -G and S is 85.15.
Warning : Angle already quite extreme.
Angle between old and new gradient is 1.25.
Current parameter steps : |dx|=0.25, max(dxi)=0.24 for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=0.81, clampings=0
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH : 52.846477729,  1.000000000,  0.298929099,  0.012564560
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 20 (173 calls) :
Current goal function and gradient : f=38.609m, reduction df=9.09m, |g|=0.35
X :  1.504083306,  0.000000000,  0.387918463, -0.516254916
G : -0.222357187,  0.000000000,  0.017868084, -0.274407338
S :  0.040425951,  0.000000000,  0.573495816,  0.078121142
Angle between -G and S is 81.15.
Angle between old and new gradient is 0.20.
Current parameter steps : |dx|=0.45, max(dxi)=0.44 for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=0.77, clampings=0
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.

Current real Hessian Matrix H  : 
116.678274,  0.000000, -8.452142, 63.838688
 0.000000,  1.000000,  0.000000,  0.000000
-8.452142,  0.000000,  1.448134, -6.223619
63.838688,  0.000000, -6.223619, 40.045789

RMS value of diagonal matrix entries : 61.69
RMS value of non-diagonal entries : 26.41
Ratio of RMS values : 2.336:1

Current real inverse Hessian Matrix : 
 0.095952,  0.000000, -0.293139, -0.198519
 0.000000,  1.000000,  0.000000,  0.000000
-0.293139,  0.000000,  2.974965,  0.929652
-0.198519,  0.000000,  0.929652,  0.485918

Current inverse Hessian Matrix from BFGS : 
 0.330955,  0.000000,  1.873976,  0.001167
 0.000000,  1.000000,  0.000000,  0.000000
 1.873976,  0.000000, 20.750987,  1.922631
 0.001167,  0.000000,  1.922631,  0.408938

Sign mismatches in matrices : 4 of 16 (25.00%)
Warning : Matrix quite bad, although already many iterations.
Normalized average entry difference in matrices : 51.10%
Hint : Angle between S via iH and real iH is 2.77. Length ratio is 7.44:1.
Current eigenvalues of real iH :  3.319908382,  1.000000000,  0.230413492,  0.006513344
Current eigenvalues of BFGS iH : 21.098767209,  1.000000000,  0.372070129,  0.020042054
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 21 (179 calls) :
Current goal function and gradient : f=29.198m, reduction df=9.41m, |g|=0.92
X :  1.476018389,  0.000000000,  0.380953896, -0.483398519
G :  0.891709595,  0.000000000, -0.032652373,  0.218492456
S : -0.050635287,  0.000000000, -0.012565613,  0.059280170
Angle between -G and S is 50.56.
Angle between old and new gradient is 161.70.
Current parameter steps : |dx|=43.77m, max(dxi)=32.86m for number i=4, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=0.55, clampings=0
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH : 12.616105161,  1.000000000,  0.277880568,  0.012814942
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 22 (186 calls) :
Current goal function and gradient : f=23.602m, reduction df=5.60m, |g|=0.26
X :  1.452695216,  0.000000000,  0.620825652, -0.403445703
G : -0.170960405,  0.000000000,  0.004618825, -0.192385585
S : -0.041489716,  0.000000000,  0.426709132,  0.142228485
Angle between -G and S is 75.84.
Angle between old and new gradient is 164.07.
Current parameter steps : |dx|=0.25, max(dxi)=0.24 for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=0.56, clampings=0
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH : 12.730409422,  1.000000000,  0.284173294,  0.007549512
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 23 (192 calls) :
Current goal function and gradient : f=22.281m, reduction df=1.32m, |g|=0.48
X :  1.447236721,  0.000000000,  0.586869439, -0.400012016
G :  0.456564995,  0.000000000,  0.024434695,  0.132325262
S : -0.010598447,  0.000000000, -0.065930835,  0.006666994
Angle between -G and S is 74.31.
Angle between old and new gradient is 165.95.
Current parameter steps : |dx|=34.56m, max(dxi)=33.96m for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=0.52, clampings=0
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH :  5.035222330,  1.000000000,  0.312317228,  0.006245795
Hint : Start stepsize t good already before linear search.
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 24 (197 calls) :
Current goal function and gradient : f=21.921m, reduction df=0.36m, |g|=35.03m
X :  1.436950274,  0.000000000,  0.628662329, -0.377115569
G :  0.034314270,  0.000000000, -0.006770677, -0.001979516
S : -0.010286447,  0.000000000,  0.041792891,  0.022896447
Angle between -G and S is 54.10.
Angle between old and new gradient is 7.75.
Current parameter steps : |dx|=48.75m, max(dxi)=41.79m for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=1.00, clampings=0
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH :  3.715758872,  1.000000000,  0.250795716,  0.005361374
Hint : Start stepsize t good already before linear search.
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 25 (202 calls) :
Current goal function and gradient : f=21.866m, reduction df=54.83u, |g|=96.24m
X :  1.434960557,  0.000000000,  0.645195856, -0.373348309
G :  0.078194273,  0.000000000, -0.011832023,  0.054844924
S : -0.001989717,  0.000000000,  0.016533527,  0.003767261
Angle between -G and S is 82.08.
Angle between old and new gradient is 19.13.
Current parameter steps : |dx|=17.07m, max(dxi)=16.53m for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=1.00, clampings=0
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.

Current real Hessian Matrix H  : 
198.049153,  0.000000, -6.141379, 105.195825
 0.000000,  1.000000,  0.000000,  0.000000
-6.141379,  0.000000,  1.373311, -5.492896
105.195825,  0.000000, -5.492896, 61.216541

RMS value of diagonal matrix entries : 103.65
RMS value of non-diagonal entries : 43.08
Ratio of RMS values : 2.406:1

Current real inverse Hessian Matrix : 
 0.202982,  0.000000, -0.760280, -0.417028
 0.000000,  1.000000,  0.000000,  0.000000
-0.760280,  0.000000,  3.983466,  1.663914
-0.417028,  0.000000,  1.663914,  0.882266

Current inverse Hessian Matrix from BFGS : 
 0.186835,  0.000000, -0.751567, -0.392238
 0.000000,  1.000000,  0.000000,  0.000000
-0.751567,  0.000000,  4.308658,  1.699608
-0.392238,  0.000000,  1.699608,  0.857204

Sign mismatches in matrices : 0 of 16 (0.00%)
Normalized average entry difference in matrices : 2.47%
Hint : Angle between S via iH and real iH is 0.00. Length ratio is 1.01:1.
Current eigenvalues of real iH :  4.867199523,  1.000000000,  0.197599049,  0.003915844
Current eigenvalues of BFGS iH :  5.148733189,  1.000000000,  0.198752040,  0.005211978
Hint : Start stepsize t good already before linear search.
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 26 (207 calls) :
Current goal function and gradient : f=21.863m, reduction df=3.25u, |g|=31.19m
X :  1.435048197,  0.000000000,  0.643822814, -0.373323559
G : -0.025954853,  0.000000000,  0.002633918, -0.017100013
S :  0.000087640,  0.000000000, -0.001373042,  0.000024750
Angle between -G and S is 76.76.
Angle between old and new gradient is 179.89.
Current parameter steps : |dx|=1.38m, max(dxi)=1.37m for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=1.00, clampings=0
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.
Current eigenvalues of BFGS iH :  5.187476174,  1.000000000,  0.186542094,  0.004544544
Hint : Start stepsize t good already before linear search.
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 27 (212 calls) :
Current goal function and gradient : f=21.863m, reduction df=40.42n, |g|=2.73m
X :  1.435110480,  0.000000000,  0.643322772, -0.373498887
G :  0.002454208,  0.000000000,  0.000066050,  0.001193512
S :  0.000062283,  0.000000000, -0.000500042, -0.000175329
Angle between -G and S is 84.47.
Angle between old and new gradient is 178.71.
Current parameter steps : |dx|=0.53m, max(dxi)=0.50m for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=1.00, clampings=0
Hint : Switched to double-sided gradient because : |g| is smaller 100dxss=2.00m.
Hint : Double-sided gradient formula used
Current eigenvalues of BFGS iH :  5.401671033,  1.000000000,  0.193671437,  0.004103638

Iteration 28 (222 calls) :
Current goal function and gradient : f=21.863m, reduction df=23.31n, |g|=2.67m
X :  1.435189019,  0.000000000,  0.642990554, -0.373666417
G : -0.002531561,  0.000000000, -0.000055075, -0.000840965
S :  0.000129435,  0.000000000, -0.000547503, -0.000276092
Angle between -G and S is 86.48.
Warning : Angle already quite extreme.
Angle between old and new gradient is 179.22.
Current parameter steps : |dx|=0.38m, max(dxi)=0.33m for number i=3, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=0.61, clampings=0
Hint : Switched to double-sided gradient because : |g| is smaller 100dxss=2.00m.
Hint : Double-sided gradient formula used
Current eigenvalues of BFGS iH : 11.134456539,  1.000000000,  0.138971548,  0.002324419

Iteration 29 (232 calls) :
Current goal function and gradient : f=21.863m, reduction df=26.63n, |g|=2.73m
X :  1.435245359,  0.000000000,  0.643007103, -0.373744007
G : -0.002541834,  0.000000000, -0.000074531, -0.001002212
S :  0.000033839,  0.000000000,  0.000009940, -0.000046602
Angle between -G and S is 67.44.
Angle between old and new gradient is 0.14.
Current parameter steps : |dx|=97.30u, max(dxi)=77.59u for number i=4, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=1.66, clampings=0
Hint : Switched to double-sided gradient because : |g| is smaller 100dxss=2.00m.
Hint : Double-sided gradient formula used
Current eigenvalues of BFGS iH :  4.764016818,  1.000000000,  0.179876403,  0.003090443
Hint : Start stepsize t good already before linear search.

Iteration 30 (241 calls) :
Current goal function and gradient : f=21.863m, reduction df=1.32n, |g|=0.34m
X :  1.435229840,  0.000000000,  0.643026331, -0.373716863
G :  0.000331911,  0.000000000,  0.000027292,  0.000073663
S : -0.000015519,  0.000000000,  0.000019228,  0.000027143
Angle between -G and S is 71.12.
Angle between old and new gradient is 178.77.
Current parameter steps : |dx|=36.71u, max(dxi)=27.14u for number i=4, min(dxi)=0.0 for number i=2
Current steps in linear search : i=0, step size is t=1.00, clampings=0
Hint : Switched to double-sided gradient because : Maximum step dxi smaller than dxss in gradient calculation. |g| is smaller 100dxss=2.00m.
Hint : Double-sided gradient formula used

Current real Hessian Matrix H  : 
197.360270,  0.000000, -6.220142, 104.932322
 0.000000,  1.000000,  0.000000,  0.000000
-6.220142,  0.000000,  1.380773, -5.537801
104.932322,  0.000000, -5.537801, 61.141725

RMS value of diagonal matrix entries : 103.31
RMS value of non-diagonal entries : 42.97
Ratio of RMS values : 2.404:1

Current real inverse Hessian Matrix : 
 0.199679,  0.000000, -0.745825, -0.410243
 0.000000,  1.000000,  0.000000,  0.000000
-0.745825,  0.000000,  3.923156,  1.635329
-0.410243,  0.000000,  1.635329,  0.868538

Current inverse Hessian Matrix from BFGS : 
 0.196990,  0.000000, -0.736316, -0.404123
 0.000000,  1.000000,  0.000000,  0.000000
-0.736316,  0.000000,  3.895250,  1.613484
-0.404123,  0.000000,  1.613484,  0.854622

Sign mismatches in matrices : 0 of 16 (0.00%)
Normalized average entry difference in matrices : 0.61%
Hint : Angle between S via iH and real iH is 0.00. Length ratio is 0.98:1.
Current eigenvalues of real iH :  4.789999474,  1.000000000,  0.197445235,  0.003927586
Current eigenvalues of BFGS iH :  4.745610029,  1.000000000,  0.197333796,  0.003918726
Hint : Start stepsize t good already before linear search.

Iteration 31 (250 calls) :
Current goal function and gradient : f=21.863m, reduction df=70.85f, |g|=2.83u
X :  1.435229697,  0.000000000,  0.643026919, -0.373716538
G : -0.000002293,  0.000000000,  0.000000099, -0.000001649
S : -0.000000143,  0.000000000,  0.000000589,  0.000000325
Angle between -G and S is 83.06.
Angle between old and new gradient is 172.15.
Current parameter steps : |dx|=0.69u, max(dxi)=0.59u for number i=3, min(dxi)=0.0 for number i=2
Hint : Steps xi-xi-1 already smaller than dx for gradient calculation.
Current steps in linear search : i=0, step size is t=1.00, clampings=0
Hint : Switched to double-sided gradient because : |g| is smaller than 12.00u. Maximum step dxi smaller than dxss in gradient calculation. |g| is smaller 100dxss=2.00m.
Hint : Double-sided gradient formula used
Current eigenvalues of BFGS iH :  4.760645827,  1.000000000,  0.197884780,  0.003906285

BFGS iterations : 31, f=21.863m, |g|=71.37n - gradient too small (good).
Number of warnings during optimization : 63
Function calls in gradient calculation : 161
Function calls in line search : 97
Single-sided gradient calculations : 24
Double-sided gradient calculations : 8
f=21.8629m compared to 2.5757 at start
Sum of all dx steps : 8.199 compared to direct step of 1.616 (ratio is 5.07:1)
|g|=71.3666n compared to 2.9321 at start

Final X vector :      1.435229697,  0.000000000,  0.643026919, -0.373716538
Compared to start X :  0.000000000,  0.000000000,  0.000000000,  0.000000000
X(start)-X(final) : -1.435230  0.000000 -0.643027  0.373717 
Final G vector :     -0.000000061,  0.000000000,  0.000000002, -0.000000036
Compared to start G : -1.264035203,  0.000000000,  0.875911703,  2.496403136

Real final Hessian matrix H : 
197.360567,  0.000000, -6.220132, 104.932477
 0.000000,  1.000000,  0.000000, -0.000003
-6.220132,  0.000000,  1.380760, -5.537792
104.932477, -0.000003, -5.537792, 61.141803

RMS value of diagonal matrix entries : 103.31
RMS value of non-diagonal entries : 42.97
Ratio of RMS values : 2.404:1

Real final inverse Hessian matrix iH : 
 0.199684, -0.000001, -0.745853, -0.410254
-0.000001,  1.000000,  0.000005,  0.000002
-0.745853,  0.000005,  3.923302,  1.635389
-0.410254,  0.000002,  1.635389,  0.868562

inverse H from BFGS : 
 0.198420,  0.000000, -0.740520, -0.407066
 0.000000,  1.000000,  0.000000,  0.000000
-0.740520,  0.000000,  3.903472,  1.621385
-0.407066,  0.000000,  1.621385,  0.860545

Sign mismatches in matrices : 0 of 16 (0.00%)
Normalized average entry difference in matrices : 0.38%
Hint : Angle between S via iH and real iH is 0.00. Length ratio is 0.98:1.
Eigenvalues of real final iH :  4.790174863,  1.000000000,  0.197445450,  0.003927580
Eigenvalues of final BFGS iH :  4.760645827,  1.000000000,  0.197884780,  0.003906285

Eigenvector 1 : -0.181398709,  0.000000000,  0.901210062,  0.393592343
Eigenvector 2 :  0.000000000,  1.000000000,  0.000000000,  0.000000000
Eigenvector 3 : -0.442548734,  0.000000000, -0.432224711,  0.785705045
Eigenvector 4 :  0.878205629,  0.000000000, -0.031657912,  0.477234376

Post-evaluation of optimization progress :
Step size parameter t : maximum=53.626, minimum=0.125, average=3.070

Gradient reached geometric mean of max(final value,goal)=0.10u and start value (go=2.93) at iteration 29
Function decreased already by 99% at iteration 20
Function decreased already by 99.9% at iteration 22
All parameters at better than 3% to final ones already at iteration 22
All parameters at better than 1% to final ones already at iteration 25


2nd optimization run : 
Started 2nd time within Anpass for Typ=4
Settings : Gradient limit=1.00n, maximum iterations=81, with clampings
0: pure eval, 1: goal function call, 2: goal function gradient, 3: Hessian
Warning : Used single-sided gradient calculation is maybe inaccurate for parameter2, delta(f) is already small.

Real Hessian Matrix H at start : 
197.360567,  0.000000, -6.220132, 104.932479
 0.000000,  1.000000,  0.000000,  0.000000
-6.220132,  0.000000,  1.380760, -5.537789
104.932479,  0.000000, -5.537789, 61.141809

RMS value of diagonal matrix entries : 103.31
RMS value of non-diagonal entries : 42.97
Ratio of RMS values : 2.404:1

Real inverse Hessian Matrix iH at start : 
 0.199681,  0.000000, -0.745843, -0.410249
 0.000000,  1.000000,  0.000000,  0.000000
-0.745843,  0.000000,  3.923257,  1.635367
-0.410249,  0.000000,  1.635367,  0.868551

Eigenvalues of real damped start iH :  1.000000000,  1.000000000,  1.000000000,  1.000000000
Eigenvalues of real iH at start :  4.790117134,  1.000000000,  0.197445466,  0.003927580
Eigenvalues of real H at start : 254.609683306,  5.064689620,  1.000000000,  0.208763162

Start values : f=21.863m, |g|=2.07m
X :  1.435229697,  0.000000000,  0.643026919, -0.373716538
Initial step size t=0.500
Warning : Single-sided gradient used. This might be the reason for problems in linear search.
Iteration 1 (33 calls) :
Current goal function and gradient : f=21.863m, reduction df=0.83f, |g|=2.07m
X :  1.435229697,  0.000000000,  0.643026919, -0.373716538
G :  0.001973228,  0.000000000,  0.000013810,  0.000611279
S : -0.001973228,  0.000000000, -0.000013810, -0.000611279
Angle between -G and S is -0.00.
Angle between old and new gradient is 0.00.
Current parameter steps : |dx|=0.98p, max(dxi)=0.93p for number i=1, min(dxi)=0.0 for number i=2
Hint : Steps xi-xi-1 already smaller than dx for gradient calculation.
Current steps in linear search : i=26, step size is t=0.47n, clampings=0
Warning : Quite many iterations in linear search!
Hint : Switched to double-sided gradient because : Maximum step dxi smaller than dxss in gradient calculation. |g| is smaller 100dxss=2.00m.
Hint : Double-sided gradient formula used
Current eigenvalues of BFGS iH :  1.000000000,  1.000000000,  1.000000000,  1.000000000

Warning : Probably rounding errors, large stepsize or large high-order derivatives caused problems in linear search. Quite some oscillation!Quite large -df/(t*psg)!

Iteration 2 (72 calls) :
Current goal function and gradient : f=21.863m, reduction df=-52.04a, |g|=71.54n
X :  1.435229697,  0.000000000,  0.643026919, -0.373716538
G : -0.000000062,  0.000000000,  0.000000002, -0.000000036
S : -0.000000005,  0.000000000, -0.000000002,  0.000000016
Angle between -G and S is 68.87.
Angle between old and new gradient is 177.48.
Current parameter steps : |dx|=0.0, max(dxi)=0.0 for number i=1, min(dxi)=0.0 for number i=1
Hint : Steps xi-xi-1 already smaller than dx for gradient calculation.
Current steps in linear search : i=29, step size is t=63.17E-24, clampings=0
Warning : Quite too many iterations in linear search!
Hint : Switched to double-sided gradient because : Close to end. |g| is smaller than 12.00u. Maximum step dxi smaller than dxss in gradient calculation. |g| is smaller 100dxss=2.00m.
Hint : Double-sided gradient formula used
Current eigenvalues of BFGS iH :  1.000000000,  1.000000000,  1.000000000,  0.000000000
Warning : Inverse Hessian used in BFGS has extreme eigenvalue ratio.

BFGS iterations : 2, f=21.863m, |g|=71.60n - stepsize too small.
Number of warnings during optimization : 6
Function calls in gradient calculation : 21
Function calls in line search : 59
Single-sided gradient calculations : 1
Double-sided gradient calculations : 2
f=21.8629m compared to 21.8629m at start
Sum of all dx steps : 0.978p compared to direct step of 0.978p (ratio is 1.00:1)
|g|=71.6047n compared to 2.0658m at start

Final X vector :      1.435229697,  0.000000000,  0.643026919, -0.373716538
Compared to start X :  1.435229697,  0.000000000,  0.643026919, -0.373716538
X(start)-X(final) :  0.000000  0.000000  0.000000  0.000000 
Final G vector :     -0.000000061,  0.000000000,  0.000000002, -0.000000037
Compared to start G :  0.001973228,  0.000000000,  0.000013810,  0.000611279

Real final Hessian matrix H : 
197.360561,  0.000000, -6.220142, 104.932484
 0.000000,  0.000000,  0.000000,  0.000004
-6.220142,  0.000000,  1.380765, -5.537788
104.932484,  0.000004, -5.537788, 61.141819

RMS value of diagonal matrix entries : 103.31
RMS value of non-diagonal entries : 42.97
Ratio of RMS values : 2.404:1

Real final inverse Hessian matrix iH : 
 0.199701, 14.468759, -0.745917, -0.410291
14.468759, 7687229.360946, -58.676424, -30.701058
-0.745917, -58.676424,  3.923538,  1.635525
-0.410291, -30.701058,  1.635525,  0.868639

inverse H from BFGS : 
 0.087610,  0.000000, -0.006384, -0.282654
 0.000000,  1.000000,  0.000000,  0.000000
-0.006384,  0.000000,  0.999955, -0.001978
-0.282654,  0.000000, -0.001978,  0.912435

Sign mismatches in matrices : 2 of 16 (12.50%)
Normalized average entry difference in matrices : 50.00%
Hint : Angle between S via iH and real iH is 1.54. Length ratio is 0.13:1.
Eigenvalues of real final iH : 7687229.361544177,  4.789907423,  0.197445536,  0.003927580
Eigenvalues of final BFGS iH :  1.000000000,  1.000000000,  1.000000000,  0.000000000
Warning : Inverse Hessian used in BFGS has extreme eigenvalue ratio.

Eigenvector 1 : -0.292310019,  0.000000000, -0.135837036,  0.946627251
Eigenvector 2 : -0.046523156,  0.000000000,  0.990708648,  0.127796597
Eigenvector 3 :  0.000000000,  1.000000000,  0.000000000,  0.000000000
Eigenvector 4 :  0.955191315,  0.000000000,  0.006683861,  0.295913632

Post-evaluation of optimization progress :
Step size parameter t : maximum=0.500, minimum=63.169E-24, average=0.167

Gradient reached geometric mean of max(final value,goal)=71.60n and start value (go=2.07m) at iteration 1
Function decreased already by 99% at iteration 1
Function decreased already by 99.9% at iteration 1
All parameters at better than 3% to final ones already at iteration 1
All parameters at better than 1% to final ones already at iteration 1

