A refinement indicator for classification problems based on impurity measures (e.g. More...

#include <ImpurityRefinementIndicator.hpp>

Inheritance diagram for sgpp::base::ImpurityRefinementIndicator:

Public Types
typedef GridPoint	counter_key_type

typedef std::pair< size_t, double >	value_type

Public Types inherited from sgpp::base::RefinementFunctor
typedef double	value_type

Public Member Functions
size_t	getRefinementsNum () const override
	Returns the maximal number of points that should be refined. More...

double	getRefinementThreshold () const override
	Returns the threshold for refinement. More...

	ImpurityRefinementIndicator (Grid &grid, DataMatrix &dataset, DataVector alphas, DataVector w1, DataVector *w2, DataVector &classesComputed, double threshold=0.0, size_t refinementsNum=1)
	Constructor. More...

virtual double	operator() (GridPoint &point) const
	This should be returning a refinement indicator for the specified grid point. More...

double	operator() (GridStorage &storage, size_t seq) const override
	This should be returning a refinement value for every grid point. More...

double	start () const override
	Returns the lower bound of refinement criterion (e.g., alpha or error) (lower bound). More...

void	update (GridPoint &point)
	Update normal vector of SVM. More...

Public Member Functions inherited from sgpp::base::RefinementFunctor
virtual double	getTotalRefinementValue (GridStorage &storage) const
	Returns the total sum of local (error) indicators used for refinement. More...

	RefinementFunctor ()
	Constructor. More...

virtual	~RefinementFunctor ()
	Destructor. More...

Public Attributes
DataVector *	alphas

DataVector *	w1

DataVector *	w2

Protected Attributes
DataVector &	classesComputed

DataMatrix &	dataset

size_t	refinementsNum

double	threshold

Detailed Description

A refinement indicator for classification problems based on impurity measures (e.g.

gini impurity, entropy impurity,...). It calculates local impurities based on the information from the provided data set. If the indicator is applied within the SVM learner, the normal vector needs to be extended after each refinement.

Member Typedef Documentation

◆ counter_key_type

typedef GridPoint sgpp::base::ImpurityRefinementIndicator::counter_key_type

◆ value_type

typedef std::pair<size_t, double> sgpp::base::ImpurityRefinementIndicator::value_type

Constructor & Destructor Documentation

◆ ImpurityRefinementIndicator()

sgpp::base::ImpurityRefinementIndicator::ImpurityRefinementIndicator	(	Grid &	grid,
		DataMatrix &	dataset,
		DataVector *	alphas,
		DataVector *	w1,
		DataVector *	w2,
		DataVector &	classesComputed,
		double	threshold = `0.0`,
		size_t	refinementsNum = `1`
	)

Constructor.

Parameters

grid	The grid to refine.
dataset	The set of data points used to compute impurities
alphas	The weights corresponding to the support vectors (only required for SVM learner)
w1	Normal vector (only required for SVM learner)
w2	Normal vector computed with abs values (only required for SVM learner)
classesComputed	The predicted labels for the data points from dataset
threshold	The refinement threshold; Only grid points with indicator values greater than this threshold will be refined
refinementsNum	The max amount of grid points to be refined

Member Function Documentation

◆ getRefinementsNum()

size_t sgpp::base::ImpurityRefinementIndicator::getRefinementsNum ( ) const

overridevirtual

Returns the maximal number of points that should be refined.

The maximal number of points to refine is set in the constructor of the implementing class.

Returns: Number of points that should refined. Default value: 1.

Reimplemented from sgpp::base::RefinementFunctor.

References refinementsNum.

◆ getRefinementThreshold()

double sgpp::base::ImpurityRefinementIndicator::getRefinementThreshold ( ) const

overridevirtual

Returns the threshold for refinement.

Only the grid points with absolute value of refinement criterion greater than this threshold will be refined.

Returns: Threshold value for refinement. Default value: 0.

Implements sgpp::base::RefinementFunctor.

References threshold.

Referenced by sgpp::base::ImpurityRefinement::getIndicator().

◆ operator()() [1/2]

double sgpp::base::ImpurityRefinementIndicator::operator() ( GridPoint & point ) const

virtual

This should be returning a refinement indicator for the specified grid point.

The point with the highest value will be refined first.

Parameters

point The grid point for which to calculate an indicator value

Returns: The indicator value

References classesComputed, dataset, chess::dim, sgpp::base::DataVector::get(), sgpp::base::DataMatrix::get(), sgpp::base::HashGridPoint::getDimension(), sgpp::base::HashGridPoint::getIndex(), sgpp::base::HashGridPoint::getLevel(), sgpp::base::DataMatrix::getNrows(), h, python.statsfileInfo::i, level, sgpp::combigrid::pow(), sgpp::base::DataVector::set(), and sgpp::base::DataVector::setAll().

◆ operator()() [2/2]

double sgpp::base::ImpurityRefinementIndicator::operator()	(	GridStorage &	storage,
		size_t	seq
	)		const

overridevirtual

This should be returning a refinement value for every grid point.

The point with the highest value will be refined first.

Parameters

storage	Reference to the grids storage object
seq	Sequence number in the coefficients array

Returns: refinement value

Implements sgpp::base::RefinementFunctor.

◆ start()

double sgpp::base::ImpurityRefinementIndicator::start ( ) const

overridevirtual

Returns the lower bound of refinement criterion (e.g., alpha or error) (lower bound).

The refinement value of grid points to be refined have to be larger than this value

Returns: lower bound

Implements sgpp::base::RefinementFunctor.

◆ update()

void sgpp::base::ImpurityRefinementIndicator::update ( GridPoint & point )

Update normal vector of SVM.

For each new grid point the normal vector has to be extended by one component. Only required for SVMLearner!

Parameters

point The new grid point

References alphas, sgpp::base::DataVector::append(), dataset, chess::dim, sgpp::base::Basis< LT, IT >::eval(), sgpp::base::DataVector::get(), sgpp::base::DataMatrix::get(), sgpp::base::Grid::getBasis(), sgpp::base::HashGridPoint::getDimension(), sgpp::base::HashGridPoint::getIndex(), sgpp::base::HashGridPoint::getLevel(), sgpp::base::DataMatrix::getNrows(), level, w1, and w2.

Referenced by python.uq.refinement.RefinementStrategy.Ranking::rank(), sgpp::base::ImpurityRefinement::refineGridpointsCollection(), and python.learner.LearnedKnowledge.LearnedKnowledge::setMemento().

Member Data Documentation

◆ alphas

DataVector* sgpp::base::ImpurityRefinementIndicator::alphas

Referenced by sgpp::base::ImpurityRefinement::refineGridpointsCollection(), and update().

◆ classesComputed

DataVector& sgpp::base::ImpurityRefinementIndicator::classesComputed

protected

Referenced by operator()().

◆ dataset

DataMatrix& sgpp::base::ImpurityRefinementIndicator::dataset

protected

Referenced by operator()(), and update().

◆ refinementsNum

size_t sgpp::base::ImpurityRefinementIndicator::refinementsNum

protected

Referenced by getRefinementsNum().

◆ threshold

double sgpp::base::ImpurityRefinementIndicator::threshold

protected

Referenced by getRefinementThreshold().

◆ w1

DataVector* sgpp::base::ImpurityRefinementIndicator::w1

Referenced by sgpp::base::ImpurityRefinement::refineGridpointsCollection(), and update().

◆ w2

DataVector* sgpp::base::ImpurityRefinementIndicator::w2

Referenced by update().

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

base/src/sgpp/base/grid/generation/functors/ImpurityRefinementIndicator.hpp
base/src/sgpp/base/grid/generation/functors/ImpurityRefinementIndicator.cpp

Public Types

Public Member Functions

Public Attributes

Protected Attributes

Detailed Description

Member Typedef Documentation

◆ counter_key_type

◆ value_type

Constructor & Destructor Documentation

◆ ImpurityRefinementIndicator()

Member Function Documentation

◆ getRefinementsNum()

◆ getRefinementThreshold()

◆ operator()() [1/2]

◆ operator()() [2/2]

◆ start()

◆ update()

Member Data Documentation

◆ alphas

◆ classesComputed

◆ dataset

◆ refinementsNum

◆ threshold

◆ w1

◆ w2