Levenberg-Marquardt

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Levenberg-Marquardt

Lorenzo Isella
Dear All,
In doing some fittings I tried several R packages and, for my specific
problem, I get good results with the minpack.lm library, which
implements the Levenberg-Marquardt (LM) algorithm for nonlinear
least-square optimization.
However, the nls.lm package does not allow the user to set any constrain
(and I would need mainly some linear constrains) on the fitting
parameters.
Do you know about any package in R implementing constrains in LM?
Unfortunately, nls did not work very well with my set of data.
Cheers

Lorenzo

_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
Reply | Threaded
Open this post in threaded view
|

Re: Levenberg-Marquardt

Dirk Eddelbuettel

On 21 May 2006 at 02:10, Lorenzo Isella wrote:
| Dear All,
| In doing some fittings I tried several R packages and, for my specific
| problem, I get good results with the minpack.lm library, which
| implements the Levenberg-Marquardt (LM) algorithm for nonlinear
| least-square optimization.
| However, the nls.lm package does not allow the user to set any constrain
| (and I would need mainly some linear constrains) on the fitting
| parameters.
| Do you know about any package in R implementing constrains in LM?
| Unfortunately, nls did not work very well with my set of data.

I continue to be surprised by your apparent lack of success when it comes to
finding the already supplied documentation. Try

        > ?optim
        > help.search("optimization")

optim() may well be your friend here as it implements several algorithms
including box-constrained optimization.

Also, and now with my listmaster head on, allow me to remind you that this
list is meant to finance-related matters.  Neither this post, nor your
preceding, and repeated, posts about multi-modal density estimation are
concerned with finance per se.  

Therefore, I would suggest that you send those questions to r-help instead.

Thanks, Dirk

--
Hell, there are no rules here - we're trying to accomplish something.
                                                  -- Thomas A. Edison

_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance