[Scilab-users] Toolbox for Scilab 6

Samuel Gougeon sgougeon at free.fr
Sun Aug 5 00:05:52 CEST 2018


Le 04/08/2018 à 21:54, Claus Futtrup a écrit :
> Hi there
>
> Is there anyone who could help me with the CMA-ES optimization toolbox 
> for Scilab 6?
>
> I wasn't successful in creating a functioning toolbox for Scilab 5 in 
> the first place.

Yet, you have published some binaries...

What are the problems?
Since there is no unit test, we could expect that examples work. But 
they don't.
And they can't. Examples:

  * cma_ask, cma_new: the examples include
      o a call to cma_stop(), but there is no page for it. => Please
        document it, or remove it from the example..
      o some calls to undefined fitfun()...
  * cma_optim:
      o the first line uses getf(), that was removed from Scilab 5.3.0
        (2010-12-16..) << 5.5
      o again, fitfun is called without being defined. Now we understand
        its meaning.
  * cma_tell:
      o y(i) = costf(X(i)); // costf needs to be defined   ==> ??

If there are no unit tests, not a single complete runnable example, no 
code up-to-date with at least Scilab 5.4 (for which the first release 
was published), your question is not just about getting help to port the 
module from Scilab 5.4 to 5.5 and 6.0...

About the macros: some advices to make the code usable on Scilab 6.0 
(and 5):
Most of functions are defined in a single file: cma.sci. In Scilab 6, 
only the first function of the file (or the one having the name of the 
file) is public. Other ones are private, and can be called only by the 
public one.
So, in pratical,

  * split the file cma.sci in as many files .sci as needed. internal
    functions (not documented) that are shared by several separate
    (public or internal) functions must also be defined separately.

Cheers
Samuel


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.scilab.org/pipermail/users/attachments/20180805/02508a86/attachment.htm>


More information about the users mailing list