
<Busko 14-OCT-97>

Added Levenberg-Marquardt method.



<Busko 25-SEP-97>

Added error catching code in user function evaluation.



<Busko 11-SEP-97>

Revised spectral regions in nl_xunit. 



<Busko 11-MAR-96>

Added ccd noise model. 
Modified files: nlfit.h, nlparams.x, nlfit.x



<Busko 26-FEB-96>

Major modification to include minimization in chi-square instead of r.m.s. 
A new parameter, "sigma", was defined for the three non-linear tasks. It is 
used to set the data error bars when none are supplied in the input data. 
Also added external control for several amoeba parameters.
Modified files: nlsumsq.x, mlamoeba.x, nlfit.x, nlparams.x, nlinit.x,
nlfit.h, mkpkg



<Bushouse 14-MAY-93>

Corrected inappropriate scaling of weights when calculating error estimates for 
fit coefficients for the case where function=user and errtype=bars. Weights 
should not be scaled when function=user.       
Modified files: nlfit.x



<Busko 27-JUL-91>

Corrected wrong common allocation when working with user function. Increased 
number of replications in bootstrap resampling.
Modified files: nlfit.x, nlvector.x.



<Busko 23-JUL-90>

Modified 2-d Gaussian to include ellipticity and position angle as function 
parameters.
Modified files: nlfit.h,nl2dgauss.x,nlscale.x,nleval.x



<Busko 12-JUL-90>

The library is now able to handle functions of two variables. Added 2-dim 
Gaussian to the function repertoire.
Modified files: nleval.x, nlfit.h, nlfit.x, nlinit.x, nlparams.x, nlscale.x, 
nlsumsq.x, nlvector.x, nlzeval.x, nlxscale.x



<Busko 08-JUN-90>

Introduced baseline with linear slope in Gaussians and cgauss functions. 
Both zero point and slope are now an integral part of the function coefficients. 
"baseline" structure element was eliminated.
Modified files: nlfit.h,nleval.x,nlscale.x,nlinit.x,nlparams.x



<Busko 23-MAY-90>	

The user function is now implemented through a parser-interpreter, taken from 
stsdas$lib/xtools. Also, the constants which control the amount of simplex 
expansion/contraction can now be user-selected, in case of user function.
Modified files: nlfit.h, nluserf.com, nluserf.x, nlinit.x, nlparams.x, 
nlamoeba.x.



<Busko 5-FEB-90>

New function type: GALPROF (galaxy brigthness profile), and its components, 
GALBULGE and GALDISK, were added to the library. Parameters that control the 
amount of simplex expansion/contratction in the amoeba code were fine-tuned 
for the functional type.
Modified files: nlfit.h, nleval.x, nlscale.x, nlgalprof.x,nlamoeba.x



<Busko 30-JUL-89>

New macros were defined, in order to make access to the coefficient, flag 
and error vectors independent of the physical order of the particular 
coefficients of each function.
Modified files: nlamoeba.x, nlscale.x, nleval.x, nlfit.h



<Busko 20-JUL-89>

New function type CGAUSS (constrained Gaussians) was added to the library.
Modified files: nlfit.h, nleval.x, nlscale.x, nlamoeba.x



<Busko 20-JUL-89>

A new routine, nl_scale, was added to the library in order to take care of 
parameter scaling. This was being done before in nl_vector and nl_eval.
Modified files: nleval.x,nlvector.x,nlscale.x



<Busko 24-MAY-89>

Reallocation of x data vector was wrong; the bug manifested itself when using 
sub-samples from the original data.
Modified files: nlvector.x



<Busko 28-APR-89>

Modifications were made in these routines in order to incorporate two new 
function types: (i)  sum of two black-bodies; (ii) user-specified function. 			Modified files: nleval.x,nlvector.x,nlfit.x,nlfit.h, 

