NaiveBayesGibbs::NaiveBayesGibbs(std::string trainDir, std::string testDir, int parameter, int interval):
DIRICHLET_HYPERPARAMETER(parameter), INTERVAL(interval)
{
//get all train files
std::cout << "loading all files..." << std::endl;
numCategories_ = stasticsFiles(trainDir, true);
for (int c=0; c<numCategories_; c++)
categoryFiles_.push_back(0);
stasticsFiles(testDir, false);
std::cout << "load "<< trainFiles_.size() << " files for train..." << std::endl;
std::cout << "load "<< testFiles_.size() << " files for test..." << std::endl;
std::cout << "num categories : " << numCategories_ << std::endl;
//init sampling labels;
std::cout << "init sampling labels..." << std::endl;
initLabels();
//init sampling theta
std::cout << "init sampling theta..." << std::endl;
stasticsWords();
updateTheta(0);
const int numTrainFiles = trainFiles_.size();
const double Z = (numTrainFiles + numCategories_*DIRICHLET_HYPERPARAMETER -1);
logLabelNorm = log(Z);
}
NaiveBayesModel NaiveBayesGibbs::train(int numIter)
{
double accuracy2 = checkMLE();
std::cout << "start training..." << std::endl;
for (int t=1; t<=numIter; t++)//number iteration
{
std::cout << Utils::getTime() << " update theta" << std::endl;
updateTheta(t);
double accuracy = check();
std::cout << Utils::getTime() << " Iteratation num " << t << " Accuracy " << accuracy << " maximum likelihood " << accuracy2 << std::endl;
std::cout << "dirichlet hyperparameter : " << DIRICHLET_HYPERPARAMETER << std::endl;
}
return NaiveBayesModel(numCategories_, labelNames_, thetaHistory_);
}
void updateTheta(int *seed,
const int *nTry,
int *nAccept,
const double *epsilon,
double *theta,
const int *Q,
const int *G,
const double *phi,
const double *lambda) {
unsigned int seedU = (unsigned int) *seed;
updateTheta(&seedU,*nTry,nAccept,*epsilon,theta,*Q,*G,phi,lambda);
*seed = (int) seedU;
return;
}
/*!
Track the ellipse in the image I.
\param I : Image in which the ellipse appears.
*/
void
vpMeEllipse::track(const vpImage<unsigned char> &I)
{
vpCDEBUG(1) <<"begin vpMeEllipse::track()"<<std::endl ;
static int iter =0 ;
// 1. On fait ce qui concerne les ellipse (peut etre vide)
{
}
//vpDisplay::display(I) ;
// 2. On appelle ce qui n'est pas specifique
{
try{
vpMeTracker::track(I) ;
}
catch(...)
{
vpERROR_TRACE("Error caught") ;
throw ;
}
// std::cout << "number of signals " << numberOfSignal() << std::endl ;
}
// 3. On revient aux ellipses
{
// Estimation des parametres de la droite aux moindres carre
suppressPoints() ;
setExtremities() ;
try{
leastSquare() ; }
catch(...)
{
vpERROR_TRACE("Error caught") ;
throw ;
}
seekExtremities(I) ;
setExtremities() ;
try
{
leastSquare() ;
}
catch(...)
{
vpERROR_TRACE("Error caught") ;
throw ;
}
// suppression des points rejetes par la regression robuste
suppressPoints() ;
setExtremities() ;
//reechantillonage si necessaire
reSample(I) ;
// remet a jour l'angle delta pour chaque point de la liste
updateTheta() ;
computeMoments();
// Remise a jour de delta dans la liste de site me
if (vpDEBUG_ENABLE(2))
{
display(I,vpColor::red) ;
vpMeTracker::display(I) ;
vpDisplay::flush(I) ;
}
// computeAngle(iP1, iP2) ;
//
// if (iter%5==0)
// {
// sample(I) ;
// try{
// leastSquare() ; }
// catch(...)
// {
// vpERROR_TRACE("Error caught") ;
// throw ;
// }
// computeAngle(iP1, iP2) ;
// }
// seekExtremities(I) ;
//
// vpMeTracker::display(I) ;
// // vpDisplay::flush(I) ;
//
// // remet a jour l'angle theta pour chaque point de la liste
// updateTheta() ;
}
iter++ ;
vpCDEBUG(1) << "end vpMeEllipse::track()"<<std::endl ;
}
