[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Help-gsl] Comments/questions about BFGS algorithm implemented for t
Re: [Help-gsl] Comments/questions about BFGS algorithm implemented for the GSL
Mon, 24 Apr 2006 20:37:24 +0100
Alan W. Irwin writes:
> For now, though, it is already good enough to solve the non-trivial
> Rosenbrock case in a much more efficient manner than the equivalent GSL
> routine due to the good quality of the Fletcher line search algorithm.
> Finally, I will be happy to consult with anybody who does the equivalent GSL
> C code as I gain experience with robustness tweaks of the Fletcher
> line-search algorithm which may be required when substantial significance
> loss is present.
Thanks for the code! The improved algorithm is well worth adding.
If Douglas McKee is interested it should be possible to do it as a
drop-in extension so that old and new algorithms can be compared side
by side at runtime: http://www.gnu.org/software/gsl/#extensions
I've ordered a copy of Fletcher's book and will take another look at
this when it arrives.
Network Theory Ltd,
Publishing the GSL Manual - http://www.network-theory.co.uk/gsl/manual/