home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math.stat
- Path: sparky!uunet!spool.mu.edu!agate!doc.ic.ac.uk!syma!mppu3
- From: mppu3@syma.sussex.ac.uk (Conor McMenamin)
- Subject: Minimization for non-linear LSQ
- Message-ID: <1993Jan7.112313.163@syma.sussex.ac.uk>
- Organization: University of Sussex
- X-Newsreader: Tin 1.1 PL5
- Date: Thu, 7 Jan 1993 11:23:13 GMT
- Lines: 20
-
- Hi,
- I hope this isn't the wrong place to post. I am working on a
- project where I have to do non-linear least-squares fitting with large
- data sets to a complicated function. Up until now I have been using the
- Downhill Simplex Method of Nelder and Mead (see Numerical Recipes) to
- perform the minimization. This has given me good results, but is
- extremely processor-time intensive. I have tried using Powell's method
- (also see Numerical Recipes) but found that the Downhill Simplex is
- better at finding deeper 'global' minima-Powell's tends to find a stable
- local minimum and stay there. However, you may have noticed my problem,
- I only have Numerical Recipes as a source. Does anyone have any
- suggestions of references which will give a selection of minimization
- techniques I can try which are a little more state-of-the-art than those
- in NR? I am presuming that minimization techniques have emerged
- which are not mentioned in NR which may be suitable.
-
- Yours Hopefully,
- Conor.
-
- P.S. I'm only an experimental physicist, so please be gentle!
-