Well, I was reacting to the "currently there is no planned work to port
external toolboxes by ESI" because some of the actual *decent* internal
computational stuff in Scilab (e.g. UMFpack) was initially developped as
an external toolbox (by B. Pincon). As far as optimization stuff is
concerned, the Sci_ipopt+Fmincon atoms packages should be considered by
ESI.
The problem is that some users (like me) keep trying to solve not
"common use-cases" problems with Scilab. The last two cases where I had
to use something else than optim are :
1-a non-linear optimization problem with linear inequality constraints
(not that complicated, but optim does no handle this) where I had to
switch to cfsqp (but stayed with Scilab here)
2-a difficult optimization problem where an interior-point method was
the only successful one (here I had to switch to Matlab, sorry)
I could give you other cases where some computational internals of
Scilab are a bit outdated. In the past 10 years the focus has been made
on things that are visible (interface, graphics, widgets, ...) and
recently on language internals and much more. However, I think that the
time has come to also update the stuff that would allows to do serious
applied mathematics and not just solve "common use-cases". I hope that
Scilab's future now depends on ESI is a good news.
Stéphane Mottelet
Le 03/04/2017 à 09:44, Pierre Vuillemin a écrit :
While it is not internal (and not necessarily decent at this point
^^), I am developing a toolbox for optimization
<https://github.com/pivui/sopi> in Scilab. Its aim is firstly to
provide a convenient way to formulate optimization problems.
It is currently very limited (mainly dense linear optimization
problems
<https://github.com/pivui/sopi/blob/master/demos/sopi_demo_LP.sce> +
some quadratic problems) and paused due to some porting issue with
Scilab 6...but still, it may be usable at some point in the future :).
Regards,
Pierre
Le 03.04.2017 09:30, Clément David a écrit :
Hi Stéphane,
Le lundi 03 avril 2017 à 08:53 +0200, Stéphane Mottelet a écrit :
BTW are there plans to port a *decent* and internal optimization
package ?
Why not but what do you mean by *decent* ? using an external solver ?
IMHO Optimization is a wide topic where there is a lot of different
solvers available. The current
Scilab implementation might not be sufficient for all needs but cover
some of the common use-cases.
--
Clément
_______________________________________________
dev mailing list
[email protected] <mailto:[email protected]>
http://lists.scilab.org/mailman/listinfo/dev
_______________________________________________
dev mailing list
[email protected]
http://lists.scilab.org/mailman/listinfo/dev
--
Département de Génie Informatique
EA 4297 Transformations Intégrées de la Matière Renouvelable
Université de Technologie de Compiègne - CS 60319
60203 Compiègne cedex
_______________________________________________
dev mailing list
[email protected]
http://lists.scilab.org/mailman/listinfo/dev
