Re: [eigen] Parameters in Levenberg Marquard algorithm of wrong type?

[ Thread Index | Date Index | More lists.tuxfamily.org/eigen Archives ]


Hi Thomas!

Thanks for the support. If you are busy, I can also suggest a patch and send it to you for review, since I had to make some changes already to make my interface compile.

While looking at the code, some quesions arose:

- Is there a reason why all local temporary variables (e.g temp, temp1, pnorm...) are defined as class members, instead of defining locally where needed?

- There are some expressions like e.g. pnorm / Scalar(0.1), probably because the original code defined 0.1 as static constant. Would it not improve readability to replace this by Scalar(10.) * pnorm?

Regards
Claas


On 16/03/12 20:16, orzel@xxxxxxxxxxxxxxx wrote:



Mmh, indeed, i've never considered complex with those. You probably
right, I'll check about patching this

Thomas

On 15/03/2012 18:34, "Claas H. Köhler" wrote:
Hi!

Has anybody ever tried to use the LevenbergMarquardt class of the
unsupported NonLinearOptimization module with complex values? I am
asking, because all parameters specified in the Parameters struct are
of type Scalar, which would result in std::complex<double>  values.
However, if I am not mistaken, these values should be real quantities,
even if the underlying problem is complex.

Furthermore comparisons of the form
if (parameters.ftol<  0.) {...}
are used, while operator<  should not be defined for complex values at
all.

Would it not be preferable to use NumTraits<Scalar>::Real instead for
the parameters?


Regards
Claas





--
Deutsches Zentrum für Luft- und Raumfahrt e.V. (DLR)
Institut für Methodik der Fernerkundung | Experimentelle Verfahren | Münchner Str | 82234 Weßling

Claas H. Köhler
Telefon 08153 28-1274 | Telefax 08153 28-1337 | claas.koehler@xxxxxx

www.DLR.de/EOC



Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/