Re: fsolve features

[Thomas Walter]

>>>>> "John" == John W Eaton <address@hidden> writes:

    John> On  5-Jul-1999, Rolf Fabian <address@hidden> wrote a number of
    John> complaints about Octave's fsolve function including

    John>   * fsolve is not able to compute complex-valued solutions.

    John>   * fsolve discards the imaginary part of complex-valued initial 
guesses.

    John>   * fsolve discards the imaginary part of complex-valued function
    John>     evaluations.

    John> Note that warnings about discarding the imaginary part of a complex
    John> value are given by default, but may be suppressed depending on the
    John> setting of the built-in variable ok_to_lose_imaginary_part.

In some cases it helps to write the function in a way to return two
real solutions instead of a  one complex solution. Something similar
can be done for the parameters, but ...

    John> The problem is that fsolve is not intended to handle complex-valued
    John> functions or find complex-valued solutions.  It is based on Minpack's
    John> hybrd function, which is only good finding real-valued solutions for
    John> real-valued functions.

... adding support for complex values would be a big step forward.

And what about the other methods from minpack?  For example the
interfaces to some least square fitting function which are derived
from LEvenberg-Marqurdt methods?  Having fitting functions would be
really nice!  But I don't speak fortran.

    John> If you have a better solution, please consider providing it as an
    John> alternative (submit it to octave-sources, for starters).  If it is
    John> really better, then perhaps it can become the default method for
    John> fsolve.

    John> Thanks,

    John> jwe

Any hints to solve and fit complex valued function are welcome.

Bye
Thomas


-- 
Platzangst:
Der dauerhafte Zustand eines Luftballons.

----------------------------------------------
Dipl. Phys. Thomas Walter
Inst. f. Physiklische Chemie II
Egerlandstr. 3                          Tel.: ++9131-85 27326 / 27330
91058 Erlangen, Germany                 email: address@hidden



---------------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.  To ensure
that development continues, see www.che.wisc.edu/octave/giftform.html
Instructions for unsubscribing: www.che.wisc.edu/octave/archive.html
---------------------------------------------------------------------