MathGroup Archive 2001

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: FindMinimum and least square minimization

  • To: mathgroup at smc.vnet.net
  • Subject: [mg30807] Re: FindMinimum and least square minimization
  • From: john payne <j.f.b.payne at tesco.net>
  • Date: Wed, 19 Sep 2001 00:16:49 -0400 (EDT)
  • Organization: Tesco ISP
  • References: <9nam8i$nqd$1@smc.vnet.net> <9ncftv$ptr$1@smc.vnet.net>
  • Sender: owner-wri-mathgroup at wolfram.com

Mike Yukish wrote:

> Also, Find Minimum uses a steepest descent algorithm, which generally can be
> improved upon.

By default FindMinimum uses steepest descent, but you can set the option Method to
QuasiNewton, Newton, or Levenberg-Marquardt (maybe more).  I'd've thought
Levenberg-Marquardt would be the natural choice for a least-squares problem (like
Dr Satherley's).

Also the scaling of variables in FindMinimum does not always seem to work, so you
may need to scale them yourself.



  • Prev by Date: Speed issue with MathLink
  • Next by Date: FortranDef & Mathematica 4.0
  • Previous by thread: Re: FindMinimum and least square minimization
  • Next by thread: creating a palette for hilberts matrix