[scilab-Users] Academic question

[Nicolas FEZANS]

Hi!

Well I agree that NDcost is not very optimal and we could do better...
but I would not remove it from Scilab because it is just extremely
simple for the user to call optim(NDcost(... )) and works quite good
for easy optimization problems. Part of the philosophy behind software
like scilab or matlab is to provide very simple tools for engineers,
and to give them the possibility to go a lot further if required. So I
would simply let the users choose whether they want to use NDcost or
not and possibly give them enough information in the documentation in
order them to make the right choice.
Note that I am of course talking here about the functionality provided
by NDcost, not about the way it is implemented presently.

Nicolas



On Wed, Jan 26, 2011 at 8:16 AM, Michaël Baudin
<michael.baudin at scilab.org> wrote:
> Hi,
>
> Le 25/01/2011 20:26, Nicolas FEZANS a écrit :
>>
>> you may let Scilab approximate them
>> numerically thanks to the "NDcost" function.
>
> I do not suggest to use NDcost, which should not have been integrated into
> Scilab anyway.
>
> Indeed, this macro has only 2 lines where the first one calls f, and the
> second one calls derivative. We can customize the cost function directly
> ourselves, which has the major advantage of being able to fully customize
> the options of the derivative function. Moreover, we can use the full
> potential of the printing options provided by the optim function, which is
> not possible with NDcost.
>
> Best regards,
>
> Michaël Baudin
>
> --
> Michaël Baudin
> Ingénieur de développement
> michael.baudin at scilab.org
> -------------------------
> Consortium Scilab - Digiteo
> Domaine de Voluceau - Rocquencourt
> B.P. 105 - 78153 Le Chesnay Cedex
> Tel. : 01 39 63 56 87 - Fax : 01 39 63 55 94
>
>
>