Re: [Scilab-users] 'fsolver'

2016-01-27 Thread fujimoto2005 
The feature of my f(x) defined for x>0 is as follows.
f(x)<0 for x0 for x1=x2

'fsolver' gives some x such as x>x2 as a solution.
I want to get x1 as a solution. 




--
View this message in context: 
http://mailinglists.scilab.org/fsolver-tp4033340p4033345.html
Sent from the Scilab users - Mailing Lists Archives mailing list archive at 
Nabble.com.
___
users mailing list
users@lists.scilab.org
http://lists.scilab.org/mailman/listinfo/users

Re: [Scilab-users] 'fsolver'

2016-01-27 Thread Stéphane Mottelet 

Le 27/01/2016 16:54, fujimoto2005 a écrit :

The feature of my f(x) defined for x>0 is as follows.
f(x)<0 for x0 for x1=x2

'fsolver' gives some x such as x>x2 as a solution.
I want to get x1 as a solution.




--
View this message in context: 
http://mailinglists.scilab.org/fsolver-tp4033340p4033345.html
Sent from the Scilab users - Mailing Lists Archives mailing list archive at 
Nabble.com.
___
users mailing list
users@lists.scilab.org
http://lists.scilab.org/mailman/listinfo/users
A quick and dirty way of finding the minimum norm solution is to use 
lsqrsolve and to penalize the residual with the squared norm of the 
solution, like in this example (find one of the two closest to zero 
solution of cos(x)=0) :


function  res=f1(x)
  res=cos(x)
endfunction

function  res=f2(x, m)
  res=[f1(x)
   %eps^(1/4)*x]
endfunction

format(20)

[x1,v,info]=fsolve(0,f1)

disp(x1)

[x2,v,info]=lsqrsolve(0,f2,2)
[x1,v,info]=fsolve(x2,f1)

disp(x1)


But the main problem is the tuning of the penalization parameter. But when it 
works well, you can then start from x2, hoping that you are in the basin of 
attraction of your solution.


S.

--
Département de Génie Informatique
EA 4297 Transformations Intégrées de la Matière Renouvelable
Université de Technologie de Compiègne -  CS 60319
60203 Compiègne cedex

___
users mailing list
users@lists.scilab.org
http://lists.scilab.org/mailman/listinfo/users

Re: [Scilab-users] 'fsolver'

2016-01-27 Thread Serge Steer 
May be you can  use optim instead of fsolve using f(x)^2+alpha*norm(x)^2 
as cost function with alpha small enough
If you cannot compute the gradient of the cost function  you can use the 
numderivative function to estimate it.

Serge
Le 27/01/2016 16:54, fujimoto2005 a écrit :

The feature of my f(x) defined for x>0 is as follows.
f(x)<0 for x0 for x1=x2

'fsolver' gives some x such as x>x2 as a solution.
I want to get x1 as a solution.




--
View this message in context: 
http://mailinglists.scilab.org/fsolver-tp4033340p4033345.html
Sent from the Scilab users - Mailing Lists Archives mailing list archive at 
Nabble.com.
___
users mailing list
users@lists.scilab.org
http://lists.scilab.org/mailman/listinfo/users



___
users mailing list
users@lists.scilab.org
http://lists.scilab.org/mailman/listinfo/users

Re: [Scilab-users] 'fsolver'

2016-01-27 Thread fujimoto2005 
Hi, Christophe

I give a very small value as an initial point because it is known as the
character of my actual problem that this small value is necessarily smaller
than the smallest solution.
But 'fsolver' gives the largest solution which is most apart from the
initial point.

Best regards.

 



--
View this message in context: 
http://mailinglists.scilab.org/fsolver-tp4033340p4033343.html
Sent from the Scilab users - Mailing Lists Archives mailing list archive at 
Nabble.com.
___
users mailing list
users@lists.scilab.org
http://lists.scilab.org/mailman/listinfo/users

Re: [Scilab-users] optim

2016-01-27 Thread fujimoto2005 
Hi,mottelet 

Now I understand I don't need to provide the value of 'Ind '.
As explained in the help page, 'the ind input argument is a message sent
from the solver to the cost function ' .

Also I understand  I have to provide the gradient explicitly even if I don't
need its value.
Unless an user does't  provide the gradient explicitly, the solver must
estimate it numerically even if there is an analytical form of the gradient. 
In general numerical estimation  is time-consuming compared with the  the
analytical form.
So solver asks a user to provide the gradient explicitly  to avoid
time-consuming estimation.
  
Best regards



--
View this message in context: 
http://mailinglists.scilab.org/optim-tp4033329p409.html
Sent from the Scilab users - Mailing Lists Archives mailing list archive at 
Nabble.com.
___
users mailing list
users@lists.scilab.org
http://lists.scilab.org/mailman/listinfo/users

[Scilab-users] 'fsolver'

2016-01-27 Thread fujimoto2005 
if f(x)=0 has multiple solutions and I want to get the  smallest solution, is
there any way to get a such solution by 'fsolver'?
In my actual problem, 'fsolver' give the largest solution.

Best regards



--
View this message in context: 
http://mailinglists.scilab.org/fsolver-tp4033340.html
Sent from the Scilab users - Mailing Lists Archives mailing list archive at 
Nabble.com.
___
users mailing list
users@lists.scilab.org
http://lists.scilab.org/mailman/listinfo/users

Re: [Scilab-users] 'fsolver'

2016-01-27 Thread Dang Ngoc Chan, Christophe 
Hello,

> De : fujimoto2005
> Envoyé : mercredi 27 janvier 2016 15:24
>
> if f(x)=0 has multiple solutions and I want to get the  smallest solution, is 
> there any way to get a such solution by 'fsolver'?
> In my actual problem, 'fsolver' give the largest solution.

I'm afraid the solution you get depends on the initial guest.
The help page says it uses the Powell hybrid method; I don't know it, but the 
Wikipedia page for the Powell method tells it finds a *local* minimum
https://en.wikipedia.org/wiki/Powell%27s_method

So you probably have to seed with several initial guests (randomly or 
systematically) and select the smallest one (unless you have another method to 
get closer to the final solution).

HTH

--
Christophe Dang Ngoc Chan
Mechanical calculation engineer
This e-mail may contain confidential and/or privileged information. If you are 
not the intended recipient (or have received this e-mail in error), please 
notify the sender immediately and destroy this e-mail. Any unauthorized 
copying, disclosure or distribution of the material in this e-mail is strictly 
forbidden.
___
users mailing list
users@lists.scilab.org
http://lists.scilab.org/mailman/listinfo/users

Re: [Scilab-users] 'fsolver'

2016-01-27 Thread Dang Ngoc Chan, Christophe 
Of course, read "guess" instead of "guest" (-:

--
Christophe Dang Ngoc Chan
Mechanical calculation engineer


-Message d'origine-
De : Dang Ngoc Chan, Christophe
Envoyé : mercredi 27 janvier 2016 16:00
À : 'Users mailing list for Scilab' 
Objet : RE: [Scilab-users] 'fsolver'

Hello,

> De : fujimoto2005
> Envoyé : mercredi 27 janvier 2016 15:24
>
> if f(x)=0 has multiple solutions and I want to get the  smallest solution, is 
> there any way to get a such solution by 'fsolver'?
> In my actual problem, 'fsolver' give the largest solution.

I'm afraid the solution you get depends on the initial guest.
The help page says it uses the Powell hybrid method; I don't know it, but the 
Wikipedia page for the Powell method tells it finds a *local* minimum 
https://en.wikipedia.org/wiki/Powell%27s_method

So you probably have to seed with several initial guests (randomly or 
systematically) and select the smallest one (unless you have another method to 
get closer to the final solution).

HTH

--
Christophe Dang Ngoc Chan
Mechanical calculation engineer
This e-mail may contain confidential and/or privileged information. If you are 
not the intended recipient (or have received this e-mail in error), please 
notify the sender immediately and destroy this e-mail. Any unauthorized 
copying, disclosure or distribution of the material in this e-mail is strictly 
forbidden.
___
users mailing list
users@lists.scilab.org
http://lists.scilab.org/mailman/listinfo/users