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

TNMIN, IDL function minimizer

Greetings again folks!

I have put a general function minimizer called TNMIN on my web page:


TNMIN will attempt to minimize a function with respect to a set of
parameters.  How does it compare to MPFIT, also on my web page?  MPFIT
solves the least-squares problem: it allows you to perform curve
fitting.  TNMIN will seeks the minimum of a more general function, not
necessarily involved in curve fitting.  Perhaps optimization is a
better buzzword.

Better yet, TNMIN has all the bells that MPFIT does.  Namely, it
allows you fix and unfix paramter values, and to place simple bounding
constraints on parameters (ie, upper and lower limits).  MPFIT and
TNMIN have both recently gained the ability to tie parameter values to
each other.  They both share (virtually) the same calling interface to
save you from larnin' the same thing twice.

TNMIN uses a gradient technique. That means (1) that you *DO* have to
compute the gradient of your function (bummer I know); and (2) it
generally converges faster and with fewer function computations than a
simplex routine such as POWELL.  It's also better than CONSTRAINED_MIN
in the sense that the source code is visible for you to see, and it's
available for all versions of IDL.

I am swamped with work right now; if somebody can come up with a clean
numerical derivative computation, I'd gladly include it.

Happy optimization!


P.S.  Credit: TNMIN is based on TN.F by Stephen Nash

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