Re: multidimensional root-finding problem :-(

[Slaven Peles <peles at cns dot physics dot gatech dot edu>]

On May 23, 2003 07:15, Kenneth Geisshirt wrote:
> On Thu, 22 May 2003, Slaven Peles wrote:
> > If anyone on the Jlist has experience with GSL multidimensional
> > root-finding functions please help. My code is below.
>
> <disclaimer>I have no experience with the multidim solver</disclainer>
>
> Isn't the index for the Jacobian wrong?
>

The Jacobian is same like in a working example in GSL Reference Manual (which 
is also used for testing of GSL multiroots, I guess). I just copied Jacobian 
from rosenbrock_df to rosenbrock_fdf. So instead of calling rosenbrock_f and 
rosenbrock_df from rosenbrock_fdf, like in the example in the Manual, I put 
all of their content into rosenbrock_fdf. The way I understand explanations 
in the Manual this should make no difference. Yet, my program returns the 
same initial value for every iteration. 

> Moreover, the function rosenbrock_fdf is only called once.
>

Pointers for all three rosenbrocks are passed to  gsl_multiroot_fdfsolver_set 
function, and later I think gsl_multiroot_fdfsolver_iterate calls them. I 
tried to understand how does this work by loking at GSL sources, but with 
little success so far.

> Kneth

Thanks,
Slaven