Roberto Alsina wrote:
> 
> > I tried some polynomials I couldn't get it stuck in a cubic, which
> > indicates it has at least some smarts for getting out of/avoiding rel.
> > extrema. I did manage to get it stuck in a sixth order polynomial
> > which indicates it isn't that smart. (Numerical solutions for finding
> > the roots of any (finite) ordered polynomial to arbitrary precision
> > exist).
> 
> Maybe it uses multiple methods?
> 

For a general function, IMO, the best thing is to try the Newton method
and if it doesnīt find anything in a number of steps, then try another
method that isnīt so fast but is more solid, such as the Regula Falsi one.

If you know the function is a polynomial, then you can use special methods
to find the roots. Thereīs even the Sturm method that tells you the number
of roots between two points (without obtaining them).


-- 
Antonio Larrosa Jimenez
antlarr@arrakis.es       larrosa@kde.org
http://www.arrakis.es/~rlarrosa/index.html
Klein bottles for rent -- inquire within