[RBFOpt] Set initial points in RBFOpt

[Chuong Thaidoan]

Dear Giacomo,
Thank you for your advice. I studied this section and I was able to
increase the iterations by using "max_iterations=5000 and
max_evaluations=4500". Can I ask whether  your algorithm can run with a
given initial point or a number of initial points? If it is possible, how
to set an initial point?
Best regards,
Thai Doan Chuong

On Wed, Sep 30, 2020 at 11:49 AM Giacomo Nannicini <giacomo.n at gmail.com>
wrote:

> The user manual has a section on "limits and tolerances". Please refer
> to that section.
>
> G
>
> On Tue, Sep 29, 2020 at 9:02 PM Chuong Thaidoan
> <chuongthaidoan at gmail.com> wrote:
> >
> > Dear Giacomo,
> > I am running your RBFOpt method for my function but it just finished at
> 1000 evaluations. Following your manual documentation, I set
> "max_evaluations=5000" but it does not affect. Could you please let me know
> how to increase the number of evaluations as my data has more than ten
> thousand points?
> > Thank you in advance for your comments.
> > Best regards,
> > TD Chuong
> >
> > On Fri, Sep 25, 2020 at 10:02 AM Chuong Thaidoan <
> chuongthaidoan at gmail.com> wrote:
> >>
> >> Dear Giacomo,
> >> Thank you very much for your advice. I will study these suggestions.
> >> Best regards,
> >> TD Chuong
> >>
> >> On Thu, Sep 24, 2020 at 11:40 PM Giacomo Nannicini <giacomo.n at gmail.com>
> wrote:
> >>>
> >>> There's no way to do that with RBFOpt. I also don't know of any
> >>> algorithm for black-box problems that does that type of computation.
> >>> In principle, after running the optimization, say, via an object "alg"
> >>> of class RbfoptAlgorithm, you can loop through alg.all_node_pos and
> >>> alg.all_node_val to see all points explored by the algorithm.The first
> >>> data structure contains the x coordinates, the second data structure
> >>> contains the corresponding objective function value.
> >>>
> >>> G
> >>>
> >>> On Thu, Sep 24, 2020 at 9:20 AM Chuong Thaidoan
> >>> <chuongthaidoan at gmail.com> wrote:
> >>> >
> >>> > Dear Giacomo,
> >>> > Thank you for your reply. Solutions that I mean are best solutions.
> Since some problem may have several best solutions, and  so I want to know
> how to print out all such best solutions, not just one.
> >>> > Best regards,
> >>> > TD Chuong
> >>> >
> >>> > On Thu, Sep 24, 2020 at 11:00 PM Giacomo Nannicini <
> giacomo.n at gmail.com> wrote:
> >>> >>
> >>> >> That x contains the best solution found. This is an unconstrained
> >>> >> problem so any point in the domain is a solution. I don't understand
> >>> >> what you mean by "print all the solutions".
> >>> >>
> >>> >> G
> >>> >>
> >>> >> On Thu, Sep 24, 2020 at 8:46 AM Chuong Thaidoan
> >>> >> <chuongthaidoan at gmail.com> wrote:
> >>> >> >
> >>> >> > Dear Giacomo,
> >>> >> > Thank you for your email. Yes, I mean that I would like to know
> all solutions because if I print(x), it shows only one but the problem may
> have multiple solutions. How can I print all solutions?
> >>> >> > Best regards,
> >>> >> > TD Chuong
> >>> >> >
> >>> >> > On Thu, Sep 24, 2020 at 9:58 PM Giacomo Nannicini <
> giacomo.n at gmail.com> wrote:
> >>> >> >>
> >>> >> >> After running
> >>> >> >>
> >>> >> >> val, x, itercount, evalcount, fast_evalcount = alg.optimize()
> >>> >> >>
> >>> >> >> the solution is in x and the corresponding objective is in val.
> >>> >> >>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/rbfopt/attachments/20201007/9914e401/attachment.html>