Re: [Dclib-devel] Root-finding algorithm in DLib

[Zhoulai <ze...@gm...>]

Hello,

Is there an easy way to use dlib's BOBYQA algorithm with 'long double'
precision? In the document,

http://dlib.net/dlib/optimization/optimization_bobyqa_abstract.h.html#find_min_bobyqa

it seems  that the objective function has to be  of type double vector ->
double, but I have a function of type long double vector -> long double and
its minimum has to be evaluated with long-double precision.  Do you think I
can use dlib for this task?

Thanks for your ideas.

--Zhoulai


On Sat, May 31, 2014 at 9:48 AM, Zhoulai <ze...@gm...> wrote:

> Hello,
>
> I am new to Dlib,  but I am surprised that  there is no root-finding
> procedures in Dlib. Although  the functionality provided by the
> optmization.h should also be able to solve root-finding problems, I guess
> that comes with a higher complexity?
>
> So what if I want to solve a problem as follows using Dlib?
>
> I have a one dimensional function 'f' which is not necessarily
> differentiable. I want to find a root 'r' that is closest  to an initial
> guess 'g' so that f(r)=0 ?
>
> More formally, I am looking for a low-cost way to  minimize |r-g|
> under the constraint f(r)=0;
>
> In my problem setting, it is likely the initial guess is the root, thus I
> would use 'g' as an initial guess. However, I do not have an interval [a,b]
> such that f(a).f(b)<0 which is a common request for most root-finding
> procedures.
>
> Of course the problem above can be handled by the optimization approaches
> like find_min_single_variable etc, but my guess is that it can also treated
> as a root-finding problem which is less expensive.  What do you think?
>
> Thank you for your ideas.
>
> Sncerely,
> Zhoulai
>
>
>
>
>