Re: scipy.optimize.minimize for sparse matrices

--===============0481606417120695321==
Content-Type: multipart/alternative; boundary="0000000000003707f605f303af52"

--0000000000003707f605f303af52
Content-Type: text/plain; charset="UTF-8"

On Tue, Jan 24, 2023 at 1:37 AM simon zhang <[email protected]> wrote:

> I am trying to minimize a cost function that consists of sparse matrices
> using scipy.optimize.minimze and it seems like scipy.optimize.minimize
> isn't designed for sparse matrices operations. Does anyone know any other
> functions that allow me to minimize a function made up of sparse matrices
> and vectors? Or does anyone have any suggestion how to approach my problem?
>

We would need some more details about your cost function to provide
pointers. `minimize()` doesn't particularly care about what is inside of
the cost function, so it certainly _allows_ you to use sparse matrices
inside of it. I don't think anything about its design prevents you from
using sparse matrices in any way. However, the `minimize()` solvers are
better _tuned_ for some problems rather than others, and maybe that's what
you are getting at. It is true that the `minimize()` solvers are geared for
small-dimensional, non-convex problems. If you have to use sparse matrices,
you are probably working with high-dimensional parameter spaces. If you are
constructing your cost function just out of sparse matrices and vectors, it
seems likely to me that you have a convex cost function. If so, then you
will probably want to look at various convex optimization packages:

https://www.cvxpy.org/
https://cvxopt.org/

-- 
Robert Kern

--0000000000003707f605f303af52
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div dir=3D"ltr">On Tue, Jan 24, 2023 at 1:37 AM simon zha=
ng &lt;<a href=3D"mailto:[email protected]">[email protected]</a>&gt; wrote:<br><=
/div><div class=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D"=
margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-lef=
t:1ex">I am trying to minimize a cost function that consists of sparse matr=
ices using scipy.optimize.minimze and it seems like scipy.optimize.minimize=
 isn&#39;t designed for sparse matrices operations. Does anyone know any ot=
her functions that allow me to minimize a function made up of sparse matric=
es and vectors? Or does anyone have any suggestion how to approach my probl=
em?<br></blockquote><div><br></div><div>We would need some more details abo=
ut your cost function to provide pointers. `minimize()` doesn&#39;t particu=
larly care about what is inside of the cost function, so it certainly _allo=
ws_ you to use sparse matrices inside of it. I don&#39;t think anything abo=
ut its=C2=A0design prevents you from using sparse matrices in any way. Howe=
ver, the `minimize()` solvers are better _tuned_ for some problems rather t=
han others, and maybe that&#39;s what you are getting at. It is true that t=
he `minimize()` solvers are geared for small-dimensional, non-convex proble=
ms. If you have to use sparse matrices, you are probably working with high-=
dimensional parameter spaces. If you are constructing your cost function ju=
st out of sparse matrices and vectors, it seems likely to me that you have =
a convex cost function. If so, then you will probably want to look at vario=
us convex optimization packages:</div><div><br></div><div><a href=3D"https:=
//www.cvxpy.org/">https://www.cvxpy.org/</a><br></div><div><a href=3D"https=
://cvxopt.org/">https://cvxopt.org/</a></div></div><div><br></div>-- <br><d=
iv dir=3D"ltr" class=3D"gmail_signature">Robert Kern</div></div>

--0000000000003707f605f303af52--

--===============0481606417120695321==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

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

--===============0481606417120695321==--