Re: multidi

At Sat, 3 Nov 2007 18:46:38 +0100 (CET),
alberto@xxxxxxxxxxxxxxxxxxx wrote:
> I am using the multidimensionial minimization routines, and I get very
> often an error code from gsl_multimin_fminimizer_iterate, namely
> "iteration is not making progress towards solution".
> 
> I get this error with all the gradient-based algorithms; I should say that
> the function that I minimize is computed not very accurately, in the sense
> that the gradients may not correspond too accurately to the function
> itself.
> 
> Why and when is the above mentioned error code issued?

The error code is GSL_ENOPROG - you can find the files where it is
used in the multimin/ directory.  

Roughly speaking the error code is returned if the algorithm can't
find a downhill step - maybe that occurs if the gradient is not
accurate enough.

-- 
Brian Gough