Date: May 7, 2011 10:26 PM
Subject: Re: Least square fit of nonlinear equations

   Marquardt-Levenberg-Spencer-Anselmo steepest descent.  Set the sum of all
equations to zero Be careful for multiple roots. MatLab, Mathematica and
their free/open clones all have this. Spencer worked for
Marquardt at duPont, Anselmo was Spencer's doctoral student at COlumbia.
Fifteen years ago there were some issues with matlab nonlin I don't remember.
A lot of econometric systems (rats, eviews, r-language) have nonline also.
Most theoretical mathematicianc complain nonlin methods allow folks to
develop models they don't really understand. As an undergrad (1979-81) I used
to love the nonlin in (DEC20) Mlab ( but neither I nor people I
work with can really afford the current version.

- = -
Vasos Panagiotopoulos, Columbia'81+, Reagan, Mozart, Pindus, BioStrategist
---{Nothing herein constitutes advice. Everything fully disclaimed.}---
[Homeland Security means private firearms not lazy obstructive guards]
[Urb sprawl confounds terror] [Phooey on GUI: Windows for subprime Bimbos]