2013-06-10 Patrick Alken <alken@colorado.edu>
* fdfsolver.c (gsl_multifit_fdfsolver_driver): added higher level
wrapper routine
2010-02-25 Brian Gough <bjg@network-theory.co.uk>
* lmiterate.c (iterate): changed GSL_CONTINUE to GSL_ENOPROG for
the case where the routine has made 10 or more attempts to find a
suitable trial step without success.
2009-11-25 Brian Gough <bjg@network-theory.co.uk>
* multilinear.c (gsl_multifit_linear_usvd)
(gsl_multifit_wlinear_usvd): provide an unscaled version of the
fits
2009-07-09 Brian Gough <bjg@network-theory.co.uk>
* work.c (gsl_multifit_linear_free): handle NULL argument in free
* fsolver.c (gsl_multifit_fsolver_free): handle NULL argument in
free
* fdfsolver.c (gsl_multifit_fdfsolver_free): handle NULL argument
in free
2009-06-24 Brian Gough <bjg@network-theory.co.uk>
* lmset.c (set): check the return code of fdf
2008-07-03 Brian Gough <bjg@hp2.network-theory.co.uk>
* Makefile.am (INCLUDES): use top_srcdir instead of top_builddir
2007-07-29 Brian Gough <bjg@network-theory.co.uk>
* lmset.c (set): ensure internal state is zeroed after a set
2007-01-26 Brian Gough <bjg@network-theory.co.uk>
* fsolver.c (gsl_multifit_fsolver_set): made vector argument x
const
2006-03-30 Brian Gough <bjg@network-theory.co.uk>
* fsolver.c (gsl_multifit_fsolver_alloc): minpack algorithms
require n>=p, added check
* fdfsolver.c (gsl_multifit_fdfsolver_alloc): minpack algorithms
require n>=p, added check
2006-02-20 Brian Gough <bjg@network-theory.co.uk>
* multilinear.c (gsl_multifit_linear_est): added linear estimator
* test_estimator.c (test_estimator): added test for gsl_multifit_linear_est
2005-07-03 Brian Gough <bjg@network-theory.co.uk>
* multilinear.c (gsl_multifit_linear_svd): accept a user-specified
tolerance for the SVD cutoff and return effective rank
(gsl_multifit_wlinear_svd): same
2004-12-23 Brian Gough <bjg@network-theory.co.uk>
* gsl_multifit_nlin.h: removed unused declaration of
gsl_multifit_fdjacobian
2004-06-14 Brian Gough <bjg@network-theory.co.uk>
* lmiterate.c (iterate): handle case where fnorm = 0 to avoid
division by zero
* covar.c (gsl_multifit_covar): change tolerance test to match
original code, and handle case where tolr = 0
2003-03-21 Brian Gough <bjg@network-theory.co.uk>
* lmset.c (set): removed reference to q, compute QR decomposition
in place
* lmiterate.c (iterate): removed reference to q, compute QR
decomposition in-place for R
* lmutil.c: removed compute_qtf
* lmder.c (lmder_free): removed reference to q
(lmder_alloc): removed reference to q
Tue Nov 12 22:18:14 2002 Brian Gough <bjg@network-theory.co.uk>
* lmder.c (lmder_alloc): use GSL_ERROR instead of GSL_ERROR_VAL
for internal alloc functions which return int
Thu Feb 28 20:15:33 2002 Brian Gough <bjg@network-theory.co.uk>
* lmiterate.c (iterate): return immediately if evaluation raised
error (Hans E. Plesser)
* lmpar.c (lmpar): avoid division by zero for w=0 in rank
deficient case
Mon Oct 8 19:25:55 2001 Brian Gough <bjg@network-theory.co.uk>
* test.c (main): added extra nist tests
* lmutil.c (compute_rptdx): fixed bug, permutation in rptdx vector
was incorrectly applied
* lmpar.c (compute_newton_direction): fixed bug, permutation of
newton direction vector was incorrect (should have been inverse
permutation)
Mon Jul 30 17:43:21 2001 Brian Gough <bjg@network-theory.co.uk>
* test_filip.c (test_filip): reduce tolerance on covariance test
slightly for MSVC with /O2
Sun Jul 1 22:42:34 2001 Brian Gough <bjg@network-theory.co.uk>
* multilinear.c: modified to use new-style vector views
* test_pontius.c: modified to use new-style vector views
* test_longley.c: modified to use new-style vector views
* test_fn.c: modified to use new-style vector views
* test_filip.c: modified to use new-style vector views
Tue Jun 26 21:41:30 2001 Brian Gough <bjg@network-theory.co.uk>
* test_filip.c (test_filip): reduce tolerance on covariance test
slightly
Wed Jun 20 13:11:26 2001 Brian Gough <bjg@network-theory.co.uk>
* removed unused variable work2
Tue Jun 19 23:18:01 2001 Brian Gough <bjg@network-theory.co.uk>
* multilinear.c: perform column scaling before doing fit to
improve accuracy
(gsl_multifit_linear): use modified Golub-Reinsch SVD for greater
speed
(gsl_multifit_wlinear): use modified Golub-Reinsch SVD for greater
speed
Wed Jun 6 13:32:22 2001 Brian Gough <bjg@network-theory.co.uk>
* lmder.c covar.c lmiterate.c lmset.c: updated to use new QR
calling convention (now passes workspace)
Sat Apr 28 11:46:59 2001 Brian Gough <bjg@network-theory.co.uk>
* qrsolv.c (qrsolv): removed local declaration of j to avoid
shadowing global j
Mon Apr 23 13:40:04 2001 Brian Gough <bjg@network-theory.co.uk>
* qrsolv.c (qrsolv): made function static so it is not exported
Wed Apr 18 13:39:33 2001 Brian Gough <bjg@network-theory.co.uk>
* lmpar.c (compute_newton_direction): moved final rescaling inside
loop, as in the original lmpar.f
Thu Mar 8 15:29:32 2001 Brian Gough <bjg@network-theory.co.uk>
* lmpar.c (compute_newton_direction): corrected bug that produced
negative index
Sun Feb 18 16:39:46 2001 Brian Gough <bjg@network-theory.co.uk>
* fdfsolver.c (gsl_multifit_fdfsolver_alloc): changed so that the
solver _alloc function no longer calls _set, the user must do that
separately.
Fri Sep 29 19:19:24 2000 Brian Gough <bjg@network-theory.co.uk>
* Makefile.am multifit/demo.c: removed demo from Makefile since it
is was just a temporary test.