Re: [Help-gsl] nmsimplex with constraints

[Martin Jansche]

On 11/1/06, Ines Hellmann <address@hidden> wrote:

> I am using the gsl_multimin_fminimizer with the Nelder Mead ("ambeba")
> algorithm and want to
> put some constraints on the parameters that are optimized. Like I don't
> want them to become negative.
> Is there a way to do this and if so, what is it?

You could try to transform the constrained problem to an unconstrained
one.  For example, if the original problem is expressed as a function
of x with the constraint x>0, you could let x = (y>34)? y :
log1p(exp(y)) and treat it as an unconstrained problem in y. Then x>0
by construction. For example, as y -> -inf you have x -> 0 from above.

-- mj