Re: [Help-gsl] problems with BFGS

[Sam Mason]

On 17 June 2012 13:59, Claudio Tablada <address@hidden> wrote:
> The problem is that for some values of
> the sample the algorithm converges and for other values it results in NaN.

Sounds like some sort of bug in your code; one of the more subtle
reasons for this is not keeping parameter values within their
appropriate domains.  A lot of the GSL routines will return NaN
instead of 0 when it would be useful in some situations if you're just
using them without any range checking.

As Juan said, seeing your code would help a lot!


  Sam