Re: ENH: Adding COBYQA to scipy.optimize?

--===============4662996900977107746==
Content-Type: multipart/alternative; boundary="000000000000ab3f96060f76a257"

--000000000000ab3f96060f76a257
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Sat, Jan 20, 2024 at 1:07=E2=80=AFPM RAGONNEAU, Tom [AMA] <
[email protected]> wrote:

> As discussed in https://github.com/scipy/scipy/issues/18118, COBYLA is
> the only derivative-free optimization solver available in SciPy. COBYQA i=
s
> a solver we developed with my colleague Zaikun Zhang from The Hong Kong
> Polytechnic University. It is designed to supersede COBYLA as a general
> solver.
>
> I want to propose the inclusion of the COBYQA solver in scipy.optimize. A
> Python implementation of COBYQA is available at
> https://github.com/cobyqa/cobyqa, and the related Python package is
> available on PyPI. Numerical experiments we conducted show the clear
> superiority of COBYQA over COBYLA in general.
>

Thank you for proposing this inclusion Tom. It looks promising. The
comments in #18118 are on target I think. For optimization methods we have
a set of benchmarks that show the number of successful problems solved and
number of function evaluations. If COBYQA is, as we'd expect from your
comments, shows that it is an improvement on average and/or performs
significantly better on some class of problems, then it looks like
inclusion into SciPy is a good idea.

Cheers,
Ralf



>
> I would be happy to share our thoughts on this inclusion.
>
> Cheers,
> Tom, www.tomragonneau.com.
>

--000000000000ab3f96060f76a257
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div dir=3D"ltr"><br></div><br><div class=3D"gmail_quote">=
<div dir=3D"ltr" class=3D"gmail_attr">On Sat, Jan 20, 2024 at 1:07=E2=80=AF=
PM RAGONNEAU, Tom [AMA] &lt;<a href=3D"mailto:[email protected]">t=
[email protected]</a>&gt; wrote:<br></div><blockquote class=3D"gmai=
l_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,20=
4,204);padding-left:1ex">As discussed in <a href=3D"https://github.com/scip=
y/scipy/issues/18118" rel=3D"noreferrer" target=3D"_blank">https://github.c=
om/scipy/scipy/issues/18118</a>, COBYLA is the only derivative-free optimiz=
ation solver available in SciPy. COBYQA is a solver we developed with my co=
lleague Zaikun Zhang from The Hong Kong Polytechnic University. It is desig=
ned to supersede COBYLA as a general solver.<br>
<br>
I want to propose the inclusion of the COBYQA solver in scipy.optimize. A P=
ython implementation of COBYQA is available at <a href=3D"https://github.co=
m/cobyqa/cobyqa" rel=3D"noreferrer" target=3D"_blank">https://github.com/co=
byqa/cobyqa</a>, and the related Python package is available on PyPI. Numer=
ical experiments we conducted show the clear superiority of COBYQA over COB=
YLA in general.<br></blockquote><div><br></div><div>Thank you for proposing=
 this inclusion Tom. It looks promising. The comments in #18118 are on targ=
et I think. For optimization methods we have a set of benchmarks that show =
the number of successful problems solved and number of function evaluations=
. If COBYQA is, as we&#39;d expect from your comments, shows that it is an =
improvement on average and/or performs significantly better on some class o=
f problems, then it looks like inclusion into SciPy is a good idea.</div><d=
iv><br></div><div>Cheers,<br></div><div>Ralf</div><div><br></div><div>=C2=
=A0<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px =
0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
I would be happy to share our thoughts on this inclusion.<br>
<br>
Cheers,<br>
Tom, <a href=3D"http://www.tomragonneau.com" rel=3D"noreferrer" target=3D"_=
blank">www.tomragonneau.com</a>.<br>
</blockquote></div></div>

--000000000000ab3f96060f76a257--

--===============4662996900977107746==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
SciPy-Dev mailing list -- [email protected]
To unsubscribe send an email to [email protected]
https://mail.python.org/mailman3/lists/scipy-dev.python.org/
Member address: [email protected]

--===============4662996900977107746==--