Re: [eigen] New Levenberg Marquardt stuff

[Sameer Agarwal <sameeragarwal@xxxxxxxxxx>]

Hi Thomas,

I am curious as to what was your evaluation criterion for various algorithms, and 

what were the ways in which minpack was the best.

I am asking because if  ceres is missing something along the lines of performance, numerical stability and or robustness we would like to fix it.

Thanks,

Sameer

On Fri, Dec 7, 2012 at 12:55 PM, Thomas Capricelli <orzel@xxxxxxxxxxxxxxx> wrote:

Il 07/12/2012 21:24, Clifford Yapp ha scritto:

> P.S.  As long as we're discussing LM, does
> http://joachimwuttke.de/lmfit/ look like it might be of any interest
> for Eigen?

You'll find hundreds of LM implementation on the net. By the time i
wrote the eigen stuff, the minpack one was, _by far_, the best
implementation available, so that's what i've used.

Now, i dont know Ceres yet..

Thomas

--
Thomas Capricelli <orzel@xxxxxxxxxxxxxxx>
http://www..freehackers.org/thomas/