sgpp::datadriven::MultipleClassRefinementFunctor Class Reference

Multiple class refinement is based on the zero-crossing based refinement. More...

#include <MultipleClassRefinementFunctor.hpp>

Inheritance diagram for sgpp::datadriven::MultipleClassRefinementFunctor:

Public Member Functions
double	getBorderPenalty ()
	Gets the factor, the borders are penalized with.
double	getTopPercent ()
	Gets the range in which densities are considered to be close.
	MultipleClassRefinementFunctor (std::vector< base::Grid * > grids, std::vector< base::DataVector * > alphas, std::vector< double > priors, size_t refinements_num, size_t partCombined, double thresh)
	Constructor.
double	operator() (base::GridStorage &storage, size_t seq) const override
	This should be returning a refinement value for every grid point.
void	refine ()
	Organizes the refinement of the classes uses set parameter to execute the refinement step.
void	setBorderPenalty (double newPenalty)
	Sets the factor, used to penalize the borders.
void	setTopPercent (double newPercent)
	Sets the range in which densities are considered to be close.
Public Member Functions inherited from sgpp::datadriven::ZeroCrossingRefinementFunctor
size_t	getNumGrids () override
	Returns the number of grids the functor can / does refine.
size_t	getRefinementsNum () const override
	Returns the maximal number of points that should be refined.
double	getRefinementThreshold () const override
	Returns the threshold for refinement.
double	operator() (base::GridStorage &storage, size_t seq) const override
	This should be returning a refinement value for every grid point.
void	preComputeEvaluations () override
	Precomputes grid evaluations for all grids.
void	setGridIndex (size_t grid_index) override
	Sets the index (into the vector of grids) of the grid to be refined.
double	start () const override
	Returns the lower bound of refinement criterion (e.g., alpha or error) (lower bound).
	ZeroCrossingRefinementFunctor (std::vector< base::Grid * > grids, std::vector< base::DataVector * > alphas, std::vector< double > priors, size_t refinements_num=1, bool level_penalize=false, bool pre_compute=false, double threshold=0.0)
	Constructor.
	~ZeroCrossingRefinementFunctor () override
Public Member Functions inherited from sgpp::datadriven::MultiGridRefinementFunctor
virtual	~MultiGridRefinementFunctor ()
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.
	RefinementFunctor ()
	Constructor.
virtual	~RefinementFunctor ()
	Destructor.

Additional Inherited Members
Public Types inherited from sgpp::base::RefinementFunctor
typedef double	value_type
Protected Member Functions inherited from sgpp::datadriven::ZeroCrossingRefinementFunctor
void	getChild (const base::HashGridPoint &gp, size_t d, bool left, base::HashGridPoint &child) const
std::vector< double >	getEvalVector (size_t ind, size_t seq) const
	Gets the evaluations of all grids at the coords of seq.
void	getParent (const base::HashGridPoint &gp, size_t d, base::HashGridPoint &par) const
void	goDown (base::HashGridPoint &gp, base::HashGridPoint &down, size_t d, bool left) const
	Used for non-leaf grid points.
void	goUp (base::HashGridPoint &gp, base::HashGridPoint &up, size_t d, bool left) const
	Used for leaf grid points.
bool	hasChild (const base::HashGridPoint &gp, size_t d, bool left) const
bool	isLeftChild (const base::HashGridPoint &gp, size_t d) const
int	sgn (double d) const
Protected Attributes inherited from sgpp::datadriven::ZeroCrossingRefinementFunctor
std::vector< base::DataVector * >	alphas
size_t	current_grid_index
std::vector< base::Grid * >	grids
bool	level_penalize
std::vector< std::map< std::string, double > >	pre_comp_evals
	Stores grid evaluations at all grids (vector) at the union of grid point coordinates over all grids (hashed by string representation in the map)
bool	pre_compute
std::vector< double >	priors
size_t	refinements_num
double	threshold

Detailed Description

Multiple class refinement is based on the zero-crossing based refinement.

The zero-crossings are not determinate by pairwise comparing the sings of PDFS f_1 - f_2, but by comparing the dominating class at geometric neighbors, determined for each dimension separately.

Finer levels are penalized by 2^{-|l|}.

Constructor & Destructor Documentation

MultipleClassRefinementFunctor()

sgpp::datadriven::MultipleClassRefinementFunctor::MultipleClassRefinementFunctor	(	std::vector< base::Grid * >	grids,
		std::vector< base::DataVector * >	alphas,
		std::vector< double >	priors,
		size_t	refinements_num,
		size_t	partCombined,
		double	thresh
	)

Constructor.

Parameters

grids	Vector of grids. current_grid_index specifies the grid to be refined
alphas	Vector of surpluses related to the grids
priors	Vector of priors related to the classificator
refinements_num	Maximum number of refinements done
partCombined	Number of refinement done in the combined grid
thresh	Threshold for refinement scores

Member Function Documentation

getBorderPenalty()

double sgpp::datadriven::MultipleClassRefinementFunctor::getBorderPenalty

(

)

Gets the factor, the borders are penalized with.

Returns

The border penalty

getTopPercent()

double sgpp::datadriven::MultipleClassRefinementFunctor::getTopPercent

(

)

Gets the range in which densities are considered to be close.

Gives a percentage [0,1].

Returns

Range in which densities are considered close

operator()()

double sgpp::datadriven::MultipleClassRefinementFunctor::operator() ( base::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.

References sgpp::datadriven::ZeroCrossingRefinementFunctor::current_grid_index, sgpp::base::MultipleClassPoint::getBorderScore(), sgpp::base::MultipleClassPoint::getDensity(), sgpp::base::MultipleClassPoint::getDominateClass(), sgpp::base::HashGridPoint::getLevel(), sgpp::base::MultipleClassPoint::getNeighbors(), sgpp::base::HashGridStorage::getPoint(), sgpp::base::HashGridStorage::getSequenceNumber(), sgpp::base::Grid::getStorage(), sgpp::base::MultipleClassPoint::getTopClasses(), and sgpp::datadriven::ZeroCrossingRefinementFunctor::grids.

refine()

void sgpp::datadriven::MultipleClassRefinementFunctor::refine

(

)

Organizes the refinement of the classes uses set parameter to execute the refinement step.

Creates the combined grid and starts the refinement.

References sgpp::base::Grid::getStorage(), sgpp::datadriven::ZeroCrossingRefinementFunctor::grids, refine(), sgpp::datadriven::ZeroCrossingRefinementFunctor::refinements_num, and sgpp::datadriven::ZeroCrossingRefinementFunctor::setGridIndex().

Referenced by sgpp::datadriven::ModelFittingClassification::adapt(), and refine().

setBorderPenalty()

void sgpp::datadriven::MultipleClassRefinementFunctor::setBorderPenalty

(

double

newPenalty

)

Sets the factor, used to penalize the borders.

The border score is multiplied with the factor before beeing added to the overall score.

Parameters

newPenalty

Factor to penalize the borders

setTopPercent()

void sgpp::datadriven::MultipleClassRefinementFunctor::setTopPercent

(

double

newPercent

)

Sets the range in which densities are considered to be close.

Set the percentage [0,1] the densities need to have, compared to the density of the dominating class.

Parameters

newPercent

Set the new range

---

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

• datadriven/src/sgpp/datadriven/functors/classification/MultipleClassRefinementFunctor.hpp
• datadriven/src/sgpp/datadriven/functors/classification/MultipleClassRefinementFunctor.cpp