[Ipopt] optimization with experimental data

[pierre-lin.pommier at fr.michelin.com]

Hello,

I'm looking for to solve an optimization problem. I have some experimental 
data and I want to minimize the difference (L2-norm) between the model and 
the measures.

I have some difficulties to link Ipopt solver : how can I take into 
account these experimental data in functions "eval_f", "eval_grad_f", 
"eval_g", "eval_jac_g" and "eval_h" ?

You gave an example to minimize a function without experimental data.

For instance, if my function is

y(p1, p2, x) = p1 * exp(p2 * x)

where p1 and p2 are my 2 parameters, x is my variable (the temperature for 
instance). So at each componant of x, called x_i, I must have a new point 
y(p1, p2, x_i)

the jacobian of my function is

exp(p2 * x)
p2 * p1 * exp(p2 * x)

for each x.

How can I program functions "eval_f", "eval_grad_f", "eval_g", 
"eval_jac_g" and "eval_h" ?

Thank you very much.

Sincerly,

Pierre-Lin Pommier
_______________________________________________________
Pierre-Lin POMMIER
Manufacture Française des Pneumatiques Michelin 
CTE/DTO/SIM/ET   CER  Ladoux - Bâtiment  F32 ? 2ème étage 
23 place des Carmes-Dechaux 63040 Clermont-Ferrand Cedex 9 
Tel int : 67 194                                   Fax int : 68 541 
Tel ext : (+33) (0)4 73 10 71 94      Fax ext : (+33) (0)4 73 10 85 41 
mail : pierre-lin.pommier at fr.michelin.com

Confidentialité D3
_______________________________________________________
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.coin-or.org/pipermail/ipopt/attachments/20100615/cdcfa62f/attachment.html