[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

MPFIT update (and bug notice)




Greetings--

Long time group members know that I have a small package useful for
curve fitting in IDL called MPFIT.  Generally speaking this package is
meant to incorporate all the goodness of the time-proven fortran code
called MINPACK version 1 by Jorge More', which makes it quite robust.
Upon that some extra features are added to MPFIT, such as the ability
to impose simple parameter constraints.  All well and good, right?

On March 26 I posted an "update" to the code.  I'd made some fairly
significant internal changes.  Some of the more visible changes are
improved documentation, more choices over how numerical derivatives
are computed, and overflow checking.  It also fixed an obscure bug
regarding parameter constraints.

Unfortunately I also introduced *another* typographical bug in the
handling of constraints.  The essence of the problem is that if you
impose constraints, the parameter values will almost certainly "rail"
to the constraints, and/or the fitting procedure may terminate
prematurely.  You will almost certainly notice the effect.  Somehow
this got past me.

So if you downloaded the code between March 26 and April 20, please go
and get the new version, where these problems are fixed.  With the fix
these problems have disappeared.

Craig

http://cow.physics.wisc.edu/~craigm/idl/idl.html    (under Curve Fitting)

P.S. Thanks to Rob Dimeo and Nari for bringing this to my attention.

-- 
--------------------------------------------------------------------------
Craig B. Markwardt, Ph.D.         EMAIL:    craigmnet@cow.physics.wisc.edu
Astrophysics, IDL, Finance, Derivatives | Remove "net" for better response
--------------------------------------------------------------------------