Hi Davor. I already had done what you are suggesting. In fact the example already works fine. The mistake was related to something with the NOMAD specification.
Thanks a lot for your help!
Now the example code looks like:
#include "nomad.hpp"
#include <RInside.h>
#include <Rcpp.h>
using namespace std;
/*----------------------------------------*/
/* The problem */
/*----------------------------------------*/
class My_Evaluator : public NOMAD::Evaluator {
public:
My_Evaluator ( const NOMAD::Parameters & p, RInside & R) :
NOMAD::Evaluator ( p) ,m_R(R) {}
~My_Evaluator ( void ) {}
RInside & m_R; // reference to the R instance passed to
constructor
bool eval_x ( NOMAD::Eval_Point & x ,
const NOMAD::Double & h_max ,
bool & count_eval ) const {
m_R.parseEvalQ("rFunction <- function (a,b,c){result <- (a + b + c) * 20;
return(result)}");
double x0;
x0= x[0].value();
double x1;
x1= x[1].value();
double x2;
x2= x[2].value();
m_R.assign( x0,"x0" );
m_R.assign( x1,"x1" );
m_R.assign( x2,"x2" );
m_R.parseEvalQ ("eval.data[1,1]<-x0");
m_R.parseEvalQ ("eval.data[1,2]<-x1");
m_R.parseEvalQ ("eval.data[1,3]<-x2");
m_R.parseEvalQ("predict<-rFunction(eval.data[1,1],eval.data[1,2],eval.data[1,3])");
double predict_ = m_R["predict"];
NOMAD::Double predict = predict_;
NOMAD::Double c1;
c1= x0*predict + x1+x2-72; //Only one constraint
x.set_bb_output ( 0 , x0*46.74 + x1*43.47 + x2*55.72-80); // objective
value
x.set_bb_output ( 1 , c1); // constraint 1
count_eval = true; // count a black-box evaluation
return true; // the evaluation succeeded
}
};
/*------------------------------------------*/
/* NOMAD main function */
/*------------------------------------------*/
int main ( int argc , char ** argv ) {
// display:
NOMAD::Display out ( std::cout );
out.precision ( NOMAD::DISPLAY_PRECISION_STD );
try {
RInside R(argc,argv); // create an embedded R instance
// NOMAD initializations:
NOMAD::begin ( argc , argv );
// parameters creation:
NOMAD::Parameters p ( out );
//p.set_DIMENSION (8); // number of variables
p.set_DIMENSION (3); // number of variables
vector<NOMAD::bb_output_type> bbot (2); // definition of
bbot[0] = NOMAD::OBJ; // output types
bbot[1] = NOMAD::PB;
p.set_BB_OUTPUT_TYPE ( bbot );
p.set_DISPLAY_STATS ( "bbe ( sol ) obj" );
// p.set_DISPLAY_DEGREE ( FULL_DISPLAY );
NOMAD::Point sp(3);
sp[0]=0.15;
sp[1]=0.10;
sp[2]=0.3;
p.set_X0 ( sp ); // starting point
NOMAD::Point lb ( 3 ); // Límite inferior
lb[0] = 0; //
lb[1] = 0; //
lb[2] = 0; //
p.set_LOWER_BOUND ( lb ); // all var. >= 0
NOMAD::Point ub ( 3 ); // Límite superior
ub[0] = 0.22; //
ub[1] = 0.18; //
ub[2] = 1.00; //
p.set_UPPER_BOUND ( ub );
p.set_MAX_BB_EVAL (50); // the algorithm terminates after
// 50 black-box evaluations
// parameters validation:
p.check();
// custom evaluator creation:
My_Evaluator ev ( p,R );
// algorithm creation and execution:
NOMAD::Mads mads ( p , &ev );
mads.run();
}
catch ( exception & e ) {
cerr << "\nNOMAD has been interrupted (" << e.what() << ")\n\n";
}
NOMAD::Slave::stop_slaves ( out );
NOMAD::end();
return EXIT_SUCCESS;
}
Best regards!!
César
Davor Cubranic <cubra...@stat.ubc.ca>
21/03/2011 10:33 a.m.
Para
CESAR PERNALETE <pernale...@pdvsa.com>
cc
rcpp-de...@r-forge.wu-wien.ac.at
Asunto
Re: [Rcpp-devel] R Embedded in a class function
Hi Cesar,
I don't know why your embed does not work, but I would suggest to rework
your code a bit and follow the pattern used by Dirk in his Qt example:
1. create an RInside instance "R" in the "main" function
2. pass that RInside instance by reference to the constructor for
MyEvaluator and save it to a private member variable
3. MyEvaluator::eval_x can work the same as now, except that it will begin
at the line "R.parseEvalQ" since you already have the RInside instance
Hope this helps,
Davor
----- Remitido por CESAR PERNALETE/PERNALETEC/PDV/PDVSA con fecha
23/03/2011 09:33 a.m. -----
CESAR PERNALETE/PERNALETEC/PDV/PDVSA
21/03/2011 11:23 a.m.
Para
Davor Cubranic <cubra...@stat.ubc.ca>, Dirk Eddelbuettel <e...@debian.org>
cc
rcpp-de...@r-forge.wu-wien.ac.at, rcpp-devel-boun...@r-forge.wu-wien.ac.at
Asunto
Re: [Rcpp-devel] R Embedded in a class function
Hi Davor, Dirk.
Thanks a lot for your help. I tried the way Dirk are suggesting and it
appears to work fine. The application at all is not working yet, i don't
know exactly
what is happening, so i did a consultation to the NOMAD developer. But i
believe that the problem of using the R object into a method is solved.
Please, find attached a simplified example of the code.
If you see something weird, please don't hesitate in writing to the list.
Best regards.
César
Davor Cubranic <cubra...@stat.ubc.ca>
Enviado por: rcpp-devel-boun...@r-forge.wu-wien.ac.at
17/03/2011 07:51 p.m.
Para
Dirk Eddelbuettel <e...@debian.org>
cc
rcpp-de...@r-forge.wu-wien.ac.at
Asunto
Re: [Rcpp-devel] R Embedded in a class function
On 2011-03-17, at 3:31 PM, Dirk Eddelbuettel wrote:
> | And it so happens that I just wrote an example of instantiating
RInside and
> | passing to Qt so that you'd get 'R inside Qt'. That is now in SVN for
> | RInside but I haven't even gotten around to writing a short blog post
about
> | it. Maybe on the weekend...
>
> I forgot to give the URL:
>
>
https://r-forge.r-project.org/scm/viewvc.php/pkg/inst/examples/qt/?root=rinside
>
But your code still creates R instance in "main", which it then passes to
the QtDensity instance (a subclass of QtMainWindow). OTOH, once the R
instance is stored as a member variable of QtDensity, it *is* used in
QtDensity's methods.
But I'm quibbling -- it's a good pattern to follow, so Cesar could perhaps
just adapt it to his code.
Davor
_______________________________________________
Rcpp-devel mailing list
Rcpp-devel@lists.r-forge.r-project.org
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
embedRinAmethodv2.cpp
Description: Binary data
_______________________________________________ Rcpp-devel mailing list Rcpp-devel@lists.r-forge.r-project.org https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
