sgpp::datadriven::RefinementHandler Class Reference

#include <RefinementHandler.hpp>

Public Member Functions
bool	checkReadyForRefinement () const
	Check whether all grids are consistent and the scheduler is currently allowing refinement. More...
size_t	checkRefinementNecessary (const std::string &refMonitor, size_t refPeriod, size_t batchSize, double currentValidError, double currentTrainError, size_t numberOfCompletedRefinements, RefinementMonitor &monitor, sgpp::base::AdaptivityConfiguration adaptivityConfig)
	Check whether refinement is currently necessary according to the guidelines set by the user. More...
void	doRefinementForClass (const std::string &refType, RefinementResult *refinementResult, const std::vector< std::pair< std::unique_ptr< DBMatOnlineDE >, size_t >> &onlineObjects, Grid &grid, DataVector &alpha, bool preCompute, MultiGridRefinementFunctor *refinementFunctor, size_t classIndex, sgpp::base::AdaptivityConfiguration &adaptivityConfig)
	Handles refinement for a specific class. More...
RefinementResult &	getRefinementResult (size_t classIndex)
	Fetches the currently stored refinement results for a specific class. More...
	RefinementHandler (LearnerSGDEOnOffParallel *learnerInstance, size_t numClasses)
	Creates the refinement handler for a specific learner instance. More...
void	updateClassVariablesAfterRefinement (size_t classIndex, RefinementResult *refinementResult, DBMatOnlineDE *densEst, Grid &grid)
	After refinement completes locally or refinement results have been received over MPI, this method uses the results to adjust the grid and alpha vector. More...

Protected Member Functions
size_t	handleDataAndZeroBasedRefinement (bool preCompute, MultiGridRefinementFunctor *func, size_t idx, base::Grid &grid, base::GridGenerator &gridGen) const
	Logic that handles data-based and zero-crossing refinement functors. More...
size_t	handleSurplusBasedRefinement (DBMatOnlineDE *densEst, Grid &grid, DataVector &alpha, base::GridGenerator &gridGen, sgpp::base::AdaptivityConfiguration adaptivityConfig) const
	Logic that handles surplus based refinement functors. More...

Protected Attributes
LearnerSGDEOnOffParallel *	learnerInstance
std::vector< RefinementResult >	vectorRefinementResults

Constructor & Destructor Documentation

RefinementHandler()

sgpp::datadriven::RefinementHandler::RefinementHandler	(	LearnerSGDEOnOffParallel *	learnerInstance,
		size_t	numClasses
	)

Creates the refinement handler for a specific learner instance.

Parameters

learnerInstance	The instance of the learner to handle refinement for
numClasses	The number of classes for the current problem

References learnerInstance, and vectorRefinementResults.

Member Function Documentation

checkReadyForRefinement()

bool sgpp::datadriven::RefinementHandler::checkReadyForRefinement

(

)

const

Check whether all grids are consistent and the scheduler is currently allowing refinement.

Returns

Whether refinement is currently possible

References sgpp::datadriven::LearnerSGDEOnOffParallel::checkAllGridsConsistent(), sgpp::datadriven::LearnerSGDEOnOffParallel::getScheduler(), sgpp::datadriven::MPITaskScheduler::isReadyForRefinement(), and learnerInstance.

Referenced by sgpp::datadriven::LearnerSGDEOnOffParallel::trainParallel().

checkRefinementNecessary()

size_t sgpp::datadriven::RefinementHandler::checkRefinementNecessary	(	const std::string &	refMonitor,
		size_t	refPeriod,
		size_t	batchSize,
		double	currentValidError,
		double	currentTrainError,
		size_t	numberOfCompletedRefinements,
		RefinementMonitor &	monitor,
		sgpp::base::AdaptivityConfiguration	adaptivityConfig
	)

Check whether refinement is currently necessary according to the guidelines set by the user.

Parameters

refMonitor	String constant specifying the monitor to use for refinement
refPeriod	The minimum period in which refinement cycles are allowed
batchSize	The number of instances that were added by the current batch
currentValidError	The current validation error
currentTrainError	The current training error
numberOfCompletedRefinements	The number of refinement cycles already completed
monitor	The convergence monitor, if any
adaptivityConfig	the configuration for the adaptivity of the grids

Returns

How many refinement cycles should be started

References sgpp::datadriven::LearnerSGDEOnOffParallel::getError(), sgpp::datadriven::LearnerSGDEOnOffParallel::getOffline(), sgpp::datadriven::LearnerSGDEOnOffParallel::getTrainData(), sgpp::datadriven::LearnerSGDEOnOffParallel::getValidationData(), learnerInstance, sgpp::base::AdaptivityConfiguration::numRefinements_, sgpp::datadriven::RefinementMonitor::pushToBuffer(), and sgpp::datadriven::RefinementMonitor::refinementsNecessary().

Referenced by sgpp::datadriven::LearnerSGDEOnOffParallel::trainParallel().

doRefinementForClass()

void sgpp::datadriven::RefinementHandler::doRefinementForClass	(	const std::string &	refType,
		RefinementResult *	refinementResult,
		const std::vector< std::pair< std::unique_ptr< DBMatOnlineDE >, size_t >> &	onlineObjects,
		Grid &	grid,
		DataVector &	alpha,
		bool	preCompute,
		MultiGridRefinementFunctor *	refinementFunctor,
		size_t	classIndex,
		sgpp::base::AdaptivityConfiguration &	adaptivityConfig
	)

Handles refinement for a specific class.

Parameters

refType	String constant specifying the type of refinement functor
refinementResult	The RefinementResult used to store changes for the grid
onlineObjects	The density estimation online objects
grid	the grid of the online object of the current class
alpha	the surplusses of the current class
preCompute	Whether to precompute the functor's evaluation step
refinementFunctor	The refinement functor to use
classIndex	The index of the current class for which refinement is taking place
adaptivityConfig	the configuration for the adaptivity of the grids

References sgpp::datadriven::RefinementResult::addedGridPoints, D, sgpp::datadriven::RefinementResult::deletedGridPointsIndices, sgpp::base::HashGridPoint::get(), sgpp::datadriven::LearnerSGDEOnOffParallel::getDimensionality(), sgpp::base::Grid::getGenerator(), sgpp::base::Grid::getSize(), sgpp::base::Grid::getStorage(), handleDataAndZeroBasedRefinement(), handleSurplusBasedRefinement(), python.statsfileInfo::i, learnerInstance, and updateClassVariablesAfterRefinement().

Referenced by sgpp::datadriven::LearnerSGDEOnOffParallel::doRefinementForAll().

getRefinementResult()

RefinementResult & sgpp::datadriven::RefinementHandler::getRefinementResult

(

size_t

classIndex

)

Fetches the currently stored refinement results for a specific class.

Parameters

classIndex

The class to search refinement results for

Returns

A reference to the stored refinement results

References vectorRefinementResults.

Referenced by sgpp::datadriven::LearnerSGDEOnOffParallel::computeNewSystemMatrixDecomposition(), sgpp::datadriven::LearnerSGDEOnOffParallel::doRefinementForAll(), sgpp::datadriven::LearnerSGDEOnOffParallel::mergeAlphaValues(), sgpp::datadriven::MPIMethods::receiveGridComponentsUpdate(), and sgpp::datadriven::LearnerSGDEOnOffParallel::train().

handleDataAndZeroBasedRefinement()

size_t sgpp::datadriven::RefinementHandler::handleDataAndZeroBasedRefinement ( bool  preCompute, MultiGridRefinementFunctor *  func, size_t  idx, base::Grid &  grid, base::GridGenerator &  gridGen  ) const

protected

Logic that handles data-based and zero-crossing refinement functors.

Parameters

preCompute	Whether to precompute evaluations in the functor
func	Pointer to the refinement functor itself
idx	Class index
grid	The grid for the current class
gridGen	The grid's generator for the current grid

Returns

The number of added grid points

References sgpp::base::Grid::getSize(), sgpp::datadriven::MultiGridRefinementFunctor::preComputeEvaluations(), sgpp::base::GridGenerator::refine(), and sgpp::datadriven::MultiGridRefinementFunctor::setGridIndex().

Referenced by doRefinementForClass().

handleSurplusBasedRefinement()

size_t sgpp::datadriven::RefinementHandler::handleSurplusBasedRefinement ( DBMatOnlineDE *  densEst, base::Grid &  grid, DataVector &  alpha, base::GridGenerator &  gridGen, sgpp::base::AdaptivityConfiguration  adaptivityConfig  ) const

protected

Logic that handles surplus based refinement functors.

Parameters

densEst	Online objects for use in density estimation for the current class
grid	The current classes grid
alpha	The current surpluss vector
gridGen	The current grid's grid generator
adaptivityConfig	the configuration for the adaptivity

Returns

The number of added grid points

References alpha, sgpp::op_factory::createOperationEval(), sgpp::base::DataVector::get(), sgpp::datadriven::Dataset::getDimension(), sgpp::base::HashGridStorage::getPoint(), sgpp::base::HashGridStorage::getSize(), sgpp::base::DataVector::getSize(), sgpp::base::Grid::getSize(), sgpp::base::HashGridPoint::getStandardCoordinates(), sgpp::base::Grid::getStorage(), sgpp::datadriven::LearnerSGDEOnOffParallel::getTrainData(), python.utils.sg_projections::gridStorage, learnerInstance, sgpp::base::AdaptivityConfiguration::noPoints_, friedman::p, and sgpp::base::GridGenerator::refine().

Referenced by doRefinementForClass().

updateClassVariablesAfterRefinement()

void sgpp::datadriven::RefinementHandler::updateClassVariablesAfterRefinement	(	size_t	classIndex,
		RefinementResult *	refinementResult,
		DBMatOnlineDE *	densEst,
		Grid &	grid
	)

After refinement completes locally or refinement results have been received over MPI, this method uses the results to adjust the grid and alpha vector.

If this is run on the master, the grid changes will be exported over MPI. If the system matrix is refineable, a system matrix update will be assigned to workers over MPI.

Parameters

classIndex	The index of the class being updated
refinementResult	The grid changes from the refinement cycle
densEst	A pointer to the online object specfic to this class
grid	the grid of the current density function

References sgpp::datadriven::RefinementResult::addedGridPoints, sgpp::datadriven::MPIMethods::assignSystemMatrixUpdate(), sgpp::datadriven::MPITaskScheduler::assignTaskStaticTaskSize(), sgpp::datadriven::LearnerSGDEOnOffParallel::checkGridStateConsistent(), D, sgpp::datadriven::RefinementResult::deletedGridPointsIndices, sgpp::base::HashGridStorage::deletePoints(), sgpp::datadriven::LearnerSGDEOnOffParallel::getDimensionality(), sgpp::datadriven::LearnerSGDEOnOffParallel::getLocalGridVersion(), sgpp::datadriven::LearnerSGDEOnOffParallel::getOffline(), sgpp::datadriven::LearnerSGDEOnOffParallel::getScheduler(), sgpp::base::Grid::getSize(), sgpp::base::Grid::getStorage(), sgpp::base::HashGridStorage::insert(), sgpp::datadriven::MPIMethods::isMaster(), learnerInstance, level, sgpp::base::HashGridStorage::recalcLeafProperty(), sgpp::datadriven::RECOMPUTE_SYSTEM_MATRIX_DECOMPOSITION, sgpp::datadriven::MPIMethods::sendRefinementUpdates(), sgpp::datadriven::LearnerSGDEOnOffParallel::setLocalGridVersion(), and sgpp::datadriven::LearnerSGDEOnOffParallel::updateAlpha().

Referenced by doRefinementForClass(), and sgpp::datadriven::MPIMethods::receiveGridComponentsUpdate().

Member Data Documentation

learnerInstance

LearnerSGDEOnOffParallel* sgpp::datadriven::RefinementHandler::learnerInstance

protected

Referenced by checkReadyForRefinement(), checkRefinementNecessary(), doRefinementForClass(), handleSurplusBasedRefinement(), RefinementHandler(), and updateClassVariablesAfterRefinement().

vectorRefinementResults

std::vector<RefinementResult> sgpp::datadriven::RefinementHandler::vectorRefinementResults

protected

Referenced by getRefinementResult(), and RefinementHandler().

---

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

• datadriven/src/sgpp/datadriven/application/learnersgdeonoffparallel/RefinementHandler.hpp
• datadriven/src/sgpp/datadriven/application/learnersgdeonoffparallel/RefinementHandler.cpp