[Trilinos-Users] [Trilinos-Bugs] Request of information

[Luca Heltai]

Dear Bartlett,

I'm experimenting a little with unconstrained optimization and the  
EpetraModelEvaluator, but I don't seem to be able to specify a working  
model with no constraints.

I set the InArgs as:


EpetraExt::ModelEvaluator::InArgs
SwimmerModelEvaluator::createInArgs() const
{
     InArgsSetup inArgs;
     inArgs.setModelEvalDescription(this->description());
     inArgs.set_Np(1);
     inArgs.setSupports(IN_ARG_x,false);
     return inArgs;
}

and the OutArgs as:

EpetraExt::ModelEvaluator::OutArgs
SwimmerModelEvaluator::createOutArgs() const
{
     OutArgsSetup outArgs;
     outArgs.setModelEvalDescription(this->description());
     outArgs.set_Np_Ng(1,1);
     return outArgs;
}

When I run the model, I get the following:

----------------------------------------------------
Exception on processing:
/usr/local/trilinos-8.0.5/packages/thyra/src/interfaces/nonlinear/ 
model_evaluator/ana/fundamental/Thyra_ModelEvaluatorBase.hpp:2275:

Throw number = 1

Throw test that evaluated to true: !this->supports(arg)

Thyra::ModelEvaluatorBase::OutArgs<Scalar>::assert_supports(arg):

model =  
'Thyra 
::EpetraModelEvaluator 
{epetraModel 
= 
'SwimmerModelEvaluator 
',W_factory='Thyra::AmesosLinearOpWithSolveFactory{solverType=Klu}'}':

Error, The argument arg = OUT_ARG_f is not supported!
Aborting!
----------------------------------------------------

I could not find any of the examples you mention for unconstrained  
problems....

Could you please point me to the correct documentation?

Thanks in advance.

Luca.

--
Luca Heltai <luca.heltai at gmail.com>
http://www-dimat.unipv.it/heltai
--
There are no answers, only cross references.



On 08/set/08, at 16:52, Bartlett, Roscoe A wrote:

> Michele,
>
> The MOOCHO package (Trilinos/packages/moocho) can solve general non- 
> convex nonlinear unconstrained optimization problems using a BFGS  
> line-search method.  I think the general optimization community  
> believes that BFGS is more robust and more efficient than nonlinear  
> CG in general anyway.  There are several examples in MOOCHO for  
> unconstrained problems.
>
> Cheers,
>
> - Ross