Re: [Help-gsl] Multidimensional root solver

[Andrew W. Steiner]

You can rewrite the solver iterate() function to modify the step if it
goes outside of some "allowed" region. Because I run into this
quite a bit, I've done this for O2scl
(c.f. line 780 or so of
http://o2scl.sourceforge.net/o2scl/html/gsl__mroot__hybrids_8h-source.html).
Basically, if the function fails after a proposed step, I shorten the
step by factors
of two until it does not fail anymore. However, I should also warn you that
doing this kind of thing can also make the routine converge less frequently.

HTH,
Andrew

- Andrew Steiner (c.f. O2scl at http://o2scl.sourceforge.net)
On Thu, Jul 31, 2008 at 2:42 PM, Himanshu Chandola
<address@hidden> wrote:
> Hello everyone,
>
> I've been using gsl to solve a set of polynomial equations which are non
> linear. There are n equations and n variables. And from the final solution
> i've to pick one which satisfies certain constraints. The multidimensional
> solver that i use is : gsl_multiroot_fsolver.
>
> Now, the thing with multidimensional solver is that I can't give it search
> bounds unlike the one dimensional solver. Besides this, since the
> multidimensional solver only returns me one root that it finds (the one that
> it would find first), there is no way of searching the other roots which
> would satisfy the constraints.
>
> So, I was thinking whether someone has encountered similar problem before
> and figured  solving a set of simultaneous nonlinear equations (polynomials
> for that matter) when you would want to get all of the solutions. Or at
> least the solution satisfying certain constraints (which perhaps might be
> achieved by giving search bounds to the solver, but isn't available in the
> multidimensional solver).
>
> thanks in advance
>
>
> Himanshu
>
>
> _______________________________________________
> Help-gsl mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/help-gsl
>