A monitor to decide if a learning algorithm has converged. More...

#include <RefinementMonitorConvergence.hpp>

Inheritance diagram for sgpp::datadriven::RefinementMonitorConvergence:

Public Member Functions
void	pushToBuffer (size_t numberInstances, double currentValidError, double currentTrainError) override
	Stores the current error values in the buffer. More...

	RefinementMonitorConvergence (double pDeclineThreshold, size_t pBufferSize, size_t pMinRefInterval)
	Constructor. More...

size_t	refinementsNecessary () override
	Checks if the model needs to be refined. More...

virtual	~RefinementMonitorConvergence ()
	Destructor. More...

Public Member Functions inherited from sgpp::datadriven::RefinementMonitor
virtual	~RefinementMonitor ()=default
	Destructor. More...

Detailed Description

A monitor to decide if a learning algorithm has converged.

The convergence criterion is based on the comparison of error measurements throughout the training process.

Constructor & Destructor Documentation

sgpp::datadriven::RefinementMonitorConvergence::RefinementMonitorConvergence	(	double	pDeclineThreshold,
		size_t	pBufferSize,
		size_t	pMinRefInterval
	)

Constructor.

Parameters

pDeclineThreshold	The convergence threshold
pBufferSize	Number of error measurements which are considered for convergence check
pMinRefInterval	Minimum number of iterations before next refinement is allowed to be performed

sgpp::datadriven::RefinementMonitorConvergence::~RefinementMonitorConvergence ( )

virtual

Destructor.

void sgpp::datadriven::RefinementMonitorConvergence::pushToBuffer	(	size_t	numberInstances,
		double	currentValidError,
		double	currentTrainError
	)

overridevirtual

Stores the current error values in the buffer.

If the buffer has reached the maximum size, the oldest values are removed.

Parameters

numberInstances	the number of instances that were used for the training step
currentValidError	The current validation error
currentTrainError	The current training error

size_t sgpp::datadriven::RefinementMonitorConvergence::refinementsNecessary ( )

overridevirtual

Checks if the model needs to be refined.

The convergence based monitor will at most trigger one refinement at once.

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