Re: [Bug-gsl] BFGS

[Brian Gough]

Y3YPnATOP writes:
 > 1) While I'm reading the code of BFGS I've found that implementation of BFGS 
 > method not satisfied to the theory. For example, I haven't found the Hessian 
 > matrix.

Hello,

The algorithm is Vector-BFGS.  The hessian is not computed explicitly.

 > 2) Why precision and step is constants, not vectors? My function is very 
 > complex and step should be like (10^-3, 10^-12)!!

Unfortunately that's all we have.

You can copy the file multimin/vector_bfgs.c into your application and
change the names of the functions if you want to modify the way it
works, e.g. to scale the step size differently.

-- 
regards,

Brian Gough

Network Theory Ltd,
Publishing Free Software Manuals --- http://www.network-theory.co.uk/