AdaBoostTrainer< Oracle > Class Template Reference

#include <AdaBoostTrainer.h>

Inherits AdaBoost< Oracle::ClassifierType >, and Oracle.

Public Types
typedef PatternAggregator < WeightedPattern< typename Oracle::DataType >, typename Oracle::ParamType >	AdaBoostPattern
typedef DataSetHandle < AdaBoostPattern >	DataSetType
	DataSet required by AdaBoostTrainer.
typedef Oracle::ClassifierType	WeakClassifierType
	Weak Classifier Type, reported by Oracle.
typedef BoostClassifier < WeakClassifierType >	ClassifierType
	The Final Strong Classifier as Additive Model.
typedef ClassifierType::ClassifierListType	ClassifierListType
	The internal list, inside BoostClassifier.

Public Member Functions
void	EnableMAdaBoost (bool mada)
	Set MAdaBoost Variant.
template<class SourceDataSetType >
void	SetValidationSet (const SourceDataSetType &set)
const DataSetType &	GetValidationSet () const
	return the validation Set
void	ResetValidationSet (double positive_weight=-1.0)
void	Restart (bool reOptimize)
ClassifierType &	Classifier ()
	Direct Access to Final Classifier.
const ClassifierType &	Classifier () const
	Direct Access to Final Classifier.
bool	Train ()
bool	Test ()
ReportTest	TestAndExportStat (double th, int max_concurrent_jobs)

Static Public Member Functions
static std::string	signature ()
	the internal signature of classifier (not the trainer)
Static Public Member Functions inherited from AdaBoost< Oracle::ClassifierType >
static bool	Update (const Oracle::ClassifierType &H, DataSetType &training_set, DataSetType &validation_set, bool verbose=true)

Detailed Description

template<class Oracle>
class AdaBoostTrainer< Oracle >

Implements an AdaBoost Discrete/Gentle/Real Train algorithm using a generic Oracle (the hypothesis generator) The Classifier will be BoostClassifier on a WeakClassifier type. It implements also MAdaBoost weight update rule.

Parameters

Oracle

is an hypotheis generator policy

Note

true class A positive +1, false class B negative -1

Member Function Documentation

template<class Oracle >

void AdaBoostTrainer< Oracle >::ResetValidationSet ( double  positive_weight = -1.0 )

inline

Reset weight of validation set

Parameters

positive_weight

set the weight of positive set

template<class Oracle >

void AdaBoostTrainer< Oracle >::Restart

(

bool

reOptimize

)

Reset weights and alphas and retrain the classifier

Parameters

reOptimize

change classifier params in order to deal with better response

template<class Oracle >

bool AdaBoostTrainer< Oracle >::Test

(

)

Test current network (all classifiers) on Validation Set prints the result to output console

template<class Oracle >

ReportTest AdaBoostTrainer< Oracle >::TestAndExportStat	(	double	th,
		int	max_concurrent_jobs
	)

Execute the test and report stats

template<class Oracle >

bool AdaBoostTrainer< Oracle >::Train

(

)

Performs a single training cycle Extract the best classifier from Oracle using Training Set and update the Validation Set

Note

not update the Oracle Training Set: it has been done manually

---

The documentation for this class was generated from the following file:

• include/Trainer/AdaBoostTrainer.h