Fitter object that encapsulates the usage of sparse grid density difference estimation with identity as regularization. More...

#include <ModelFittingDensityDifferenceEstimationCG.hpp>

Inheritance diagram for sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG:

Public Member Functions
bool	adapt (size_t newNoPoints, std::vector< size_t > &deletedGridPoints) override
	Performs a refinement given the new grid size and the points to coarsened.

double	computeResidual (DataMatrix &validationData) const override
	Should compute some kind of Residual to evaluate the fit of the model.

double	evaluate (const DataVector &sample) override
	Evaluate the fitted density at a single data point - requires a trained grid.

void	evaluate (DataMatrix &samples, DataVector &results) override
	Evaluate the fitted density on a set of data points - requires a trained grid.

void	fit (DataMatrix &dataset) override
	Fit the grid to the given dataset by determining the surpluses of the initial grid by the SGDE approach.

void	fit (DataMatrix &datasetP, DataMatrix &datasetQ) override
	Fit the grid to the given dataset by determining the weights of the initial grid by the SGDE approach.

void	fit (Dataset &dataset) override
	Fit the grid to the dataset by determining the weights of an initial grid.

void	fit (Dataset &datasetP, Dataset &datasetQ) override
	Fit the grid to the given dataset by determining the weights of the initial grid by the SGDE approach.

	ModelFittingDensityDifferenceEstimationCG (const FitterConfigurationDensityEstimation &config)
	Constructor.

void	reset () override
	Resets the state of the entire model.

void	resetTraining () override
	Resets any trained representations of the model, but does not reset the entire state.

void	update (DataMatrix &samples) override
	Updates the model based on new data samples (streaming, batch learning).

void	update (DataMatrix &samplesP, DataMatrix &samplesQ) override
	Updates the model based on new data samples (streaming, batch learning).

void	update (Dataset &dataset) override
	Train the grid of an existing model with new samples.

void	update (Dataset &datasetP, Dataset &datasetQ) override

void	updateRegularization (double lambda) override
	Updates the regularization parameter lambda of the underlying model.

Public Member Functions inherited from sgpp::datadriven::ModelFittingDensityEstimation
bool	adapt () override
	Improve accuracy of the fit on the given training data by adaptive refinement or coarsening of the grid and recalculate weights.

	ModelFittingDensityEstimation ()
	Default constructor.

Public Member Functions inherited from sgpp::datadriven::ModelFittingBaseSingleGrid
Grid &	getGrid ()
	Get the underlying grid object for the current model.

DataVector &	getSurpluses ()
	Get the surpluses of the current grid.

	ModelFittingBaseSingleGrid ()
	Default constructor.

	ModelFittingBaseSingleGrid (const ModelFittingBaseSingleGrid &rhs)=delete
	Copy constructor - we cannot deep copy all member variables yet.

	ModelFittingBaseSingleGrid (ModelFittingBaseSingleGrid &&rhs)=default
	Move constructor.

ModelFittingBaseSingleGrid &	operator= (const ModelFittingBaseSingleGrid &rhs)=delete
	Copy assign operator - we cannot deep copy all member variables yet.

ModelFittingBaseSingleGrid &	operator= (ModelFittingBaseSingleGrid &&rhs)=default
	Move assign operator.

std::string	storeFitter ()

virtual	~ModelFittingBaseSingleGrid ()=default
	virtual destructor.

Public Member Functions inherited from sgpp::datadriven::ModelFittingBase
Dataset *	getDataset ()

FitterConfiguration &	getFitterConfiguration ()
	Get or set the configuration of the fitter object.

const FitterConfiguration &	getFitterConfiguration () const
	Get the configuration of the fitter object.

virtual std::shared_ptr< BlacsProcessGrid >	getProcessGrid () const

	ModelFittingBase ()
	Default constructor.

	ModelFittingBase (const ModelFittingBase &rhs)=delete
	Copy constructor - we cannot deep copy all member variables yet.

	ModelFittingBase (ModelFittingBase &&rhs)=default
	Move constructor.

ModelFittingBase &	operator= (const ModelFittingBase &rhs)=delete
	Copy assign operator - we cannot deep copy all member variables yet.

ModelFittingBase &	operator= (ModelFittingBase &&rhs)=default
	Move assign operator.

virtual	~ModelFittingBase ()=default
	virtual destructor.

Additional Inherited Members
Public Attributes inherited from sgpp::datadriven::ModelFittingBase
bool	verboseSolver
	Whether the Solver produces output or not.

Protected Member Functions inherited from sgpp::datadriven::ModelFittingDensityEstimation
std::unique_ptr< sgpp::base::CoarseningFunctor >	getCoarseningFunctor ()
	Returns the refinement functor suitable for the model settings.

std::unique_ptr< sgpp::base::RefinementFunctor >	getRefinementFunctor ()
	Returns the refinement functor suitable for the model settings.

Protected Member Functions inherited from sgpp::datadriven::ModelFittingBase
Grid *	buildGrid (const sgpp::base::GeneralGridConfiguration &gridConfig) const
	Factory member function that generates a grid from configuration.

Grid *	buildGrid (const sgpp::base::GeneralGridConfiguration &gridConfig, const GeometryConfiguration &geometryConfig) const
	Factory member function that generates a grid from configuration.

SLESolver *	buildSolver (const SLESolverConfiguration &config) const
	Factory member function to build the solver for the least squares regression problem according to the config.

std::set< std::set< size_t > >	getInteractions (const GeometryConfiguration &geometryConfig)

void	reconfigureSolver (SLESolver &solver, const SLESolverConfiguration &config) const
	Configure solver based on the desired configuration.

Protected Attributes inherited from sgpp::datadriven::ModelFittingDensityEstimation
size_t	initialGridSize
	Initial number of grid points.

size_t	refinementsPerformed
	Count the amount of refinement operations performed on the current dataset.

Protected Attributes inherited from sgpp::datadriven::ModelFittingBaseSingleGrid
DataVector	alpha
	hierarchical surpluses of the grid.

std::unique_ptr< Grid >	grid
	the sparse grid that approximates the data.

Protected Attributes inherited from sgpp::datadriven::ModelFittingBase
std::unique_ptr< FitterConfiguration >	config
	Configuration object for the fitter.

Dataset *	dataset
	Pointer to sgpp::datadriven::Dataset.

Dataset *	extraDataset

std::unique_ptr< std::set< std::set< size_t > > >	interactions

std::unique_ptr< SLESolver >	solver
	Solver for the learning problem.

Detailed Description

Fitter object that encapsulates the usage of sparse grid density difference estimation with identity as regularization.

Allows usage of different grids, different solvers and different regularization techniques based on the provided configuration objects.

Constructor & Destructor Documentation

◆ ModelFittingDensityDifferenceEstimationCG()

sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG::ModelFittingDensityDifferenceEstimationCG ( const FitterConfigurationDensityEstimation & config )

explicit

Constructor.

Parameters

config configuration object that specifies grid, refinement, and regularization

References sgpp::datadriven::ModelFittingBase::config.

Member Function Documentation

◆ adapt()

bool sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG::adapt	(	size_t	newNoPoints,
		std::vector< size_t > &	deletedGridPoints
	)

overridevirtual

Performs a refinement given the new grid size and the points to coarsened.

Parameters

newNoPoints	the grid size after refinement and coarsening
deletedGridPoints	a list of indexes for grid points that will be removed

Returns: if the grid was refined (true)

Implements sgpp::datadriven::ModelFittingDensityEstimation.

References sgpp::datadriven::ModelFittingBaseSingleGrid::alpha, sgpp::base::DataVector::remove(), and sgpp::base::DataVector::resizeZero().

◆ computeResidual()

double sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG::computeResidual ( DataMatrix & validationData ) const

inlineoverridevirtual

Should compute some kind of Residual to evaluate the fit of the model.

In the case of density estimation, this is || R * alpha_lambda - b_val ||_2

This is useful for unsupervised learning models, where normal evaluation cannot be used as there are no targets.

Parameters

validationData Matrix for validation data

Returns: the residual score

Implements sgpp::datadriven::ModelFittingDensityEstimation.

◆ evaluate() [1/2]

double sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG::evaluate ( const DataVector & sample )

overridevirtual

Evaluate the fitted density at a single data point - requires a trained grid.

Parameters

sample vector with the coordinates in all dimensions of that sample.

Returns: evaluation of the trained grid.

Implements sgpp::datadriven::ModelFittingDensityEstimation.

References sgpp::datadriven::ModelFittingBaseSingleGrid::alpha, sgpp::op_factory::createOperationEval(), and sgpp::datadriven::ModelFittingBaseSingleGrid::grid.

◆ evaluate() [2/2]

void sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG::evaluate	(	DataMatrix &	samples,
		DataVector &	results
	)

overridevirtual

Evaluate the fitted density on a set of data points - requires a trained grid.

Parameters

samples	matrix where each row represents a sample and the columns contain the coordinates in all dimensions of that sample.
results	vector where each row will contain the evaluation of the respective sample on the current model.

Implements sgpp::datadriven::ModelFittingDensityEstimation.

References sgpp::datadriven::ModelFittingBaseSingleGrid::alpha, sgpp::op_factory::createOperationMultipleEval(), sgpp::base::OperationMultipleEval::eval(), and sgpp::datadriven::ModelFittingBaseSingleGrid::grid.

◆ fit() [1/4]

void sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG::fit ( DataMatrix & dataset )

inlineoverridevirtual

Fit the grid to the given dataset by determining the surpluses of the initial grid by the SGDE approach.

Requires only data samples and no targets (since those are irrelevant for the density estimation whatsoever)

Parameters

dataset the training dataset that is used to fit the model.

Implements sgpp::datadriven::ModelFittingDensityEstimation.

◆ fit() [2/4]

void sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG::fit	(	DataMatrix &	datasetP,
		DataMatrix &	datasetQ
	)

overridevirtual

Fit the grid to the given dataset by determining the weights of the initial grid by the SGDE approach.

Requires only data samples and no targets (since those are irrelevant for the density estimation whatsoever)

Parameters

datasetP	the first training dataset that is used to fit the model.
datasetQ	the second training dataset that is used to fit the model.

Implements sgpp::datadriven::ModelFittingDensityEstimation.

References sgpp::datadriven::ModelFittingBaseSingleGrid::alpha, sgpp::datadriven::ModelFittingBase::buildGrid(), sgpp::datadriven::ModelFittingBase::config, sgpp::base::DataMatrix::getNcols(), sgpp::datadriven::ModelFittingBaseSingleGrid::grid, reset(), and update().

◆ fit() [3/4]

void sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG::fit ( Dataset & dataset )

inlineoverridevirtual

Fit the grid to the dataset by determining the weights of an initial grid.

Parameters

dataset the training dataset that is used to fit the model.

Implements sgpp::datadriven::ModelFittingDensityEstimation.

◆ fit() [4/4]

void sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG::fit	(	Dataset &	datasetP,
		Dataset &	datasetQ
	)

overridevirtual

Fit the grid to the given dataset by determining the weights of the initial grid by the SGDE approach.

Parameters

datasetP	the first training dataset that is used to fit the model.
datasetQ	the second training dataset that is used to fit the model.

Implements sgpp::datadriven::ModelFittingDensityEstimation.

References sgpp::datadriven::ModelFittingBase::dataset, sgpp::datadriven::ModelFittingBase::extraDataset, fit(), and sgpp::datadriven::Dataset::getData().

Referenced by fit(), and update().

◆ reset()

void sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG::reset ( )

overridevirtual

Resets the state of the entire model.

Implements sgpp::datadriven::ModelFittingBase.

References sgpp::datadriven::ModelFittingBaseSingleGrid::grid, and sgpp::datadriven::ModelFittingDensityEstimation::refinementsPerformed.

Referenced by fit().

◆ resetTraining()

void sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG::resetTraining ( )

inlineoverridevirtual

Resets any trained representations of the model, but does not reset the entire state.

Implements sgpp::datadriven::ModelFittingDensityEstimation.

◆ update() [1/4]

void sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG::update ( DataMatrix & samples )

inlineoverridevirtual

Updates the model based on new data samples (streaming, batch learning).

Requires only the data samples and no targets (since those are irrelevant for the density estimation whatsoever)

Parameters

samples the new data samples

Implements sgpp::datadriven::ModelFittingDensityEstimation.

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

◆ update() [2/4]

void sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG::update	(	DataMatrix &	samplesP,
		DataMatrix &	samplesQ
	)

overridevirtual

Updates the model based on new data samples (streaming, batch learning).

Requires only the data samples and no targets (since those are irrelevant for the density estimation whatsoever)

Parameters

samplesP	the new data samples for first dataset
samplesQ	the new data samples for second dataset

Implements sgpp::datadriven::ModelFittingDensityEstimation.

References sgpp::base::DataVector::add(), sgpp::datadriven::ModelFittingBaseSingleGrid::alpha, sgpp::base::DataVector::componentwise_div(), sgpp::datadriven::DensityDifferenceSystemMatrix::computeUnweightedRhs(), sgpp::datadriven::ModelFittingBase::config, fit(), sgpp::base::DataMatrix::getNrows(), sgpp::datadriven::ModelFittingBaseSingleGrid::grid, sgpp::base::DataVector::mult(), sgpp::base::DataVector::setAll(), sgpp::solver::ConjugateGradients::solve(), and sgpp::base::DataVector::sub().

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

◆ update() [3/4]

void sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG::update ( Dataset & dataset )

inlineoverridevirtual

Train the grid of an existing model with new samples.

Parameters

dataset the training dataset that is used to fit the model.

Implements sgpp::datadriven::ModelFittingDensityEstimation.

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

◆ update() [4/4]

void sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG::update	(	Dataset &	datasetP,
		Dataset &	datasetQ
	)

overridevirtual

Implements sgpp::datadriven::ModelFittingDensityEstimation.

References sgpp::datadriven::ModelFittingBase::dataset, sgpp::datadriven::ModelFittingBase::extraDataset, sgpp::datadriven::Dataset::getData(), and update().

Referenced by fit(), python.uq.refinement.RefinementStrategy.Ranking::rank(), python.learner.LearnedKnowledge.LearnedKnowledge::setMemento(), and update().

◆ updateRegularization()

void sgpp::datadriven::ModelFittingDensityDifferenceEstimationCG::updateRegularization ( double lambda )

inlineoverridevirtual

Updates the regularization parameter lambda of the underlying model.

Parameters

lambda the new lambda parameter

Implements sgpp::datadriven::ModelFittingDensityEstimation.

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

datadriven/src/sgpp/datadriven/datamining/modules/fitting/ModelFittingDensityDifferenceEstimationCG.hpp
datadriven/src/sgpp/datadriven/datamining/modules/fitting/ModelFittingDensityDifferenceEstimationCG.cpp

Public Member Functions

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

◆ ModelFittingDensityDifferenceEstimationCG()

Member Function Documentation

◆ adapt()

◆ computeResidual()

◆ evaluate() [1/2]

◆ evaluate() [2/2]

◆ fit() [1/4]

◆ fit() [2/4]

◆ fit() [3/4]

◆ fit() [4/4]

◆ reset()

◆ resetTraining()

◆ update() [1/4]

◆ update() [2/4]

◆ update() [3/4]

◆ update() [4/4]

◆ updateRegularization()