comp.lang.idl-pvwave archive: archive » Re: HELP: Levenberg-Marquardt method

Home » Public Forums » archive » Re: HELP: Levenberg-Marquardt method

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

Re: HELP: Levenberg-Marquardt method [message #3673 is a reply to message #3652]

Mon, 06 March 1995 07:05

kletzing
Messages: 5
Registered: September 1994

Junior Member

I have done fitting with various types of plasma data, quite often
measured distribution functions that I wish to fit with idealized
functions. This does not usually work very well with the CURVEFIT
routine because the fit phase space, so to speak, has many local
minima. Thus if curvefit starts near a local minima it often just
finds that minima without finding other better minima. On the other
hand I have worked with time series data that have an extraneous
noise signal at one frequency that I have removed very successfully
with curvefit.

For the first type of problem, I have had more success by backing up
to less sophisticated methods such as gradient search or grid search
techniques. They take longer (often much longer) to converge, but
tend to do better when the data is not as nice. Both these techniques
are described in Bevington's book.

Could you elaborate on the error in CURVEFIT? I've never looked over
RSI's implementation very carefully.

Craig Kletzing
University of New Hampshire

Report message to a moderator

[Message index]

		Re: HELP: Levenberg-Marquardt method By: agraps on Tue, 07 March 1995 17:04
		Re: HELP: Levenberg-Marquardt method By: kletzing on Mon, 06 March 1995 07:05
		Re: HELP: Levenberg-Marquardt method By: larkum on Mon, 06 March 1995 05:48
		Re: HELP: Levenberg-Marquardt method By: agraps on Sat, 04 March 1995 09:17
		Re: HELP: Levenberg-Marquardt method By: kletzing on Fri, 03 March 1995 11:34

Previous Topic:	Re: Any Planetary Maps?
Next Topic:	Re: Help with IDL - READ_GIF!!!

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Tue Dec 02 11:07:14 PST 2025

Total time taken to generate the page: 0.48379 seconds