Detailed Description

The k-Nearest Neighbor learning algorithm.

#include <GKNN.h>

Inheritance diagram for GClasses::GKNN:

Public Types

Public Member Functions
	GKNN ()
	General-purpose constructor. More...

	GKNN (const GDomNode *pNode)
	Load from a DOM. More...

virtual	~GKNN ()

size_t	addVector (const GVec &in, const GVec &out)
	Adds a copy of pVector to the internal set. More...

void	autoTune (GMatrix &features, GMatrix &labels)
	Uses cross-validation to find a set of parameters that works well with the provided data. More...

virtual void	clear ()
	Discard any training (but not any settings) so it can be trained again. More...

void	drawRandom (size_t n)
	Specify to train by drawing 'n' random patterns from the training set. More...

GMatrix *	features ()
	Returns the internal feature set. More...

GMatrix *	labels ()
	Returns the internal label set. More...

GDistanceMetric *	metric ()
	Returns the dissimilarity metric. More...

size_t	neighborCount ()
	Returns the number of neighbors. More...

virtual void	predict (const GVec &in, GVec &out)
	See the comment for GSupervisedLearner::predict. More...

virtual void	predictDistribution (const GVec &in, GPrediction *pOut)
	See the comment for GSupervisedLearner::predictDistribution. More...

virtual GDomNode *	serialize (GDom *pDoc) const
	Marshal this object into a DOM, which can then be converted to a variety of serial formats. More...

void	setInterpolationLearner (GSupervisedLearner *pLearner, bool bTakeOwnership)
	Sets the interpolation method to "Learner" and sets the learner to use. If bTakeOwnership is true, it will delete the learner when this object is deleted. More...

void	setInterpolationMethod (InterpolationMethod eMethod)
	Sets the technique for interpolation. (If you want to use the "Learner" method, you should call SetInterpolationLearner instead of this method.) More...

void	setMetric (GDistanceMetric *pMetric, bool own)
	Sets the distance metric to use for finding neighbors. If own is true, then this object will delete pMetric when it is done with it. More...

void	setMetric (GSparseSimilarity *pMetric, bool own)
	Sets the sparse similarity metric to use for finding neighbors. If own is true, then this object will delete pMetric when it is done with it. More...

void	setNeighborCount (size_t k)
	Specify the number of neighbors to use. (The default is 1.) More...

void	setNormalizeScaleFactors (bool b)
	Specify whether to normalize the scaling of each attribute. (The default is to normalize.) More...

void	setOptimizeScaleFactors (bool b)
	If you set this to true, it will use a hill-climber to optimize the attribute scaling factors. If you set it to false (the default), it won't. More...

GSparseMatrix *	sparseFeatures ()
	Returns the internal set of sparse features. More...

virtual void	trainSparse (GSparseMatrix &features, GMatrix &labels)
	See the comment for GIncrementalLearner::trainSparse. More...

Public Member Functions inherited from GClasses::GIncrementalLearner
	GIncrementalLearner ()
	General-purpose constructor. More...

	GIncrementalLearner (const GDomNode *pNode)
	Deserialization constructor. More...

virtual	~GIncrementalLearner ()
	Destructor. More...

void	beginIncrementalLearning (const GRelation &featureRel, const GRelation &labelRel)
	You must call this method before you call trainIncremental. More...

void	beginIncrementalLearning (const GMatrix &features, const GMatrix &labels)
	A version of beginIncrementalLearning that supports data-dependent filters. More...

virtual bool	canTrainIncrementally ()
	Returns true. More...

virtual bool	isFilter ()
	Only the GFilter class should return true to this method. More...

Public Member Functions inherited from GClasses::GSupervisedLearner
	GSupervisedLearner ()
	General-purpose constructor. More...

	GSupervisedLearner (const GDomNode *pNode)
	Deserialization constructor. More...

virtual	~GSupervisedLearner ()
	Destructor. More...

void	basicTest (double minAccuracy1, double minAccuracy2, double deviation=1e-6, bool printAccuracy=false, double warnRange=0.035)
	This is a helper method used by the unit tests of several model learners. More...

virtual bool	canGeneralize ()
	Returns true because fully supervised learners have an internal model that allows them to generalize previously unseen rows. More...

void	confusion (GMatrix &features, GMatrix &labels, std::vector< GMatrix * > &stats)
	Generates a confusion matrix containing the total counts of the number of times each value was expected and predicted. (Rows represent target values, and columns represent predicted values.) stats should be an empty vector. This method will resize stats to the number of dimensions in the label vector. The caller is responsible to delete all of the matrices that it puts in this vector. For continuous labels, the value will be NULL. More...

void	precisionRecall (double *pOutPrecision, size_t nPrecisionSize, GMatrix &features, GMatrix &labels, size_t label, size_t nReps)
	label specifies which output to measure. (It should be 0 if there is only one label dimension.) The measurement will be performed "nReps" times and results averaged together nPrecisionSize specifies the number of points at which the function is sampled pOutPrecision should be an array big enough to hold nPrecisionSize elements for every possible label value. (If the attribute is continuous, it should just be big enough to hold nPrecisionSize elements.) If bLocal is true, it computes the local precision instead of the global precision. More...

const GRelation &	relFeatures ()
	Returns a reference to the feature relation (meta-data about the input attributes). More...

const GRelation &	relLabels ()
	Returns a reference to the label relation (meta-data about the output attributes). More...

double	sumSquaredError (const GMatrix &features, const GMatrix &labels, double *pOutSAE=NULL)
	Computes the sum-squared-error for predicting the labels from the features. For categorical labels, Hamming distance is used. More...

void	train (const GMatrix &features, const GMatrix &labels)
	Call this method to train the model. More...

virtual double	trainAndTest (const GMatrix &trainFeatures, const GMatrix &trainLabels, const GMatrix &testFeatures, const GMatrix &testLabels, double *pOutSAE=NULL)
	Trains and tests this learner. Returns sum-squared-error. More...

Public Member Functions inherited from GClasses::GTransducer
	GTransducer ()
	General-purpose constructor. More...

	GTransducer (const GTransducer &that)
	Copy-constructor. Throws an exception to prevent models from being copied by value. More...

virtual	~GTransducer ()

virtual bool	canImplicitlyHandleContinuousFeatures ()
	Returns true iff this algorithm can implicitly handle continuous features. If it cannot, then the GDiscretize transform will be used to convert continuous features to nominal values before passing them to it. More...

virtual bool	canImplicitlyHandleContinuousLabels ()
	Returns true iff this algorithm can implicitly handle continuous labels (a.k.a. regression). If it cannot, then the GDiscretize transform will be used during training to convert nominal labels to continuous values, and to convert nominal predictions back to continuous labels. More...

virtual bool	canImplicitlyHandleNominalFeatures ()
	Returns true iff this algorithm can implicitly handle nominal features. If it cannot, then the GNominalToCat transform will be used to convert nominal features to continuous values before passing them to it. More...

virtual bool	canImplicitlyHandleNominalLabels ()
	Returns true iff this algorithm can implicitly handle nominal labels (a.k.a. classification). If it cannot, then the GNominalToCat transform will be used during training to convert nominal labels to continuous values, and to convert categorical predictions back to nominal labels. More...

double	crossValidate (const GMatrix &features, const GMatrix &labels, size_t nFolds, double pOutSAE=NULL, RepValidateCallback pCB=NULL, size_t nRep=0, void pThis=NULL)
	Perform n-fold cross validation on pData. Returns sum-squared error. Uses trainAndTest for each fold. pCB is an optional callback method for reporting intermediate stats. It can be NULL if you don't want intermediate reporting. nRep is just the rep number that will be passed to the callback. pThis is just a pointer that will be passed to the callback for you to use however you want. It doesn't affect this method. if pOutSAE is not NULL, the sum absolute error will be placed there. More...

GTransducer &	operator= (const GTransducer &other)
	Throws an exception to prevent models from being copied by value. More...

GRand &	rand ()
	Returns a reference to the random number generator associated with this object. For example, you could use it to change the random seed, to make this algorithm behave differently. This might be important, for example, in an ensemble of learners. More...

double	repValidate (const GMatrix &features, const GMatrix &labels, size_t reps, size_t nFolds, double pOutSAE=NULL, RepValidateCallback pCB=NULL, void pThis=NULL)
	Perform cross validation "nReps" times and return the average score. pCB is an optional callback method for reporting intermediate stats It can be NULL if you don't want intermediate reporting. pThis is just a pointer that will be passed to the callback for you to use however you want. It doesn't affect this method. if pOutSAE is not NULL, the sum absolute error will be placed there. More...

virtual bool	supportedFeatureRange (double pOutMin, double pOutMax)
	Returns true if this algorithm supports any feature value, or if it does not implicitly handle continuous features. If a limited range of continuous values is supported, returns false and sets pOutMin and pOutMax to specify the range. More...

virtual bool	supportedLabelRange (double pOutMin, double pOutMax)
	Returns true if this algorithm supports any label value, or if it does not implicitly handle continuous labels. If a limited range of continuous values is supported, returns false and sets pOutMin and pOutMax to specify the range. More...

std::unique_ptr< GMatrix >	transduce (const GMatrix &features1, const GMatrix &labels1, const GMatrix &features2)
	Predicts a set of labels to correspond with features2, such that these labels will be consistent with the patterns exhibited by features1 and labels1. More...

void	transductiveConfusionMatrix (const GMatrix &trainFeatures, const GMatrix &trainLabels, const GMatrix &testFeatures, const GMatrix &testLabels, std::vector< GMatrix * > &stats)
	Makes a confusion matrix for a transduction algorithm. More...

Static Public Member Functions
static void	test ()
	Performs unit tests for this class. Throws an exception if there is a failure. More...

Static Public Member Functions inherited from GClasses::GSupervisedLearner
static void	test ()
	Runs some unit tests related to supervised learning. Throws an exception if any problems are found. More...

Protected Member Functions
virtual void	beginIncrementalLearningInner (const GRelation &featureRel, const GRelation &labelRel)
	See the comment for GIncrementalLearner::beginIncrementalLearningInner. More...

virtual bool	canImplicitlyHandleMissingFeatures ()
	See the comment for GTransducer::canImplicitlyHandleMissingFeatures. More...

size_t	findNeighbors (const GVec &vector)
	Finds the nearest neighbors of pVector. Returns the number of neighbors found. More...

void	interpolateLearner (size_t nc, const GVec &in, GPrediction pOut, GVec pOut2)
	Interpolates with the provided supervised learning algorithm. More...

void	interpolateLinear (size_t nc, const GVec &in, GPrediction pOut, GVec pOut2)
	Interpolate with each neighbor having a linear vote. (Actually it's linear with respect to the squared distance instead of the distance, because this is faster to compute.) More...

void	interpolateMean (size_t nc, const GVec &in, GPrediction pOut, GVec pOut2)
	Interpolate with each neighbor having equal vote. More...

virtual void	trainIncremental (const GVec &in, const GVec &out)
	Adds a vector to the internal set. Also, if the (k+1)th nearest neighbor of that vector is less than "elbow room" from it, then the closest neighbor is deleted from the internal set. (You might be wondering why the decision to delete the closest neighbor is determined by the distance of the (k+1)th neigbor. This enables a clump of k points to form in the most frequently sampled locations. Also, If you make this decision based on a closer neighbor, then big holes may form in the model if points are sampled in a poor order.) Call SetElbowRoom to specify the elbow room distance. More...

virtual void	trainInner (const GMatrix &features, const GMatrix &labels)
	See the comment for GSupervisedLearner::trainInner. More...

Protected Member Functions inherited from GClasses::GIncrementalLearner
virtual void	beginIncrementalLearningInner (const GMatrix &features, const GMatrix &labels)

Protected Member Functions inherited from GClasses::GSupervisedLearner
GDomNode *	baseDomNode (GDom pDoc, const char szClassName) const
	Child classes should use this in their implementation of serialize. More...

size_t	precisionRecallContinuous (GPrediction pOutput, double pFunc, GMatrix &trainFeatures, GMatrix &trainLabels, GMatrix &testFeatures, GMatrix &testLabels, size_t label)
	This is a helper method used by precisionRecall. More...

size_t	precisionRecallNominal (GPrediction pOutput, double pFunc, GMatrix &trainFeatures, GMatrix &trainLabels, GMatrix &testFeatures, GMatrix &testLabels, size_t label, int value)
	This is a helper method used by precisionRecall. More...

void	setupFilters (const GMatrix &features, const GMatrix &labels)
	This method determines which data filters (normalize, discretize, and/or nominal-to-cat) are needed and trains them. More...

virtual std::unique_ptr< GMatrix >	transduceInner (const GMatrix &features1, const GMatrix &labels1, const GMatrix &features2)
	See GTransducer::transduce. More...

Protected Attributes
bool	m_bOwnLearner

InterpolationMethod	m_eInterpolationMethod

TrainMethod	m_eTrainMethod

size_t	m_nNeighbors

bool	m_normalizeScaleFactors

bool	m_optimizeScaleFactors

bool	m_ownMetric

GKnnScaleFactorCritic *	m_pCritic

GDistanceMetric *	m_pDistanceMetric

GMatrix *	m_pFeatures

GMatrix *	m_pLabels

GSupervisedLearner *	m_pLearner

GNeighborFinderGeneralizing *	m_pNeighborFinder

GOptimizer *	m_pScaleFactorOptimizer

GSparseMatrix *	m_pSparseFeatures

GSparseSimilarity *	m_pSparseMetric

double	m_trainParam

GVec	m_valueCounts

Protected Attributes inherited from GClasses::GSupervisedLearner
GRelation *	m_pRelFeatures

GRelation *	m_pRelLabels

Protected Attributes inherited from GClasses::GTransducer
GRand	m_rand

Additional Inherited Members
Static Protected Member Functions inherited from GClasses::GSupervisedLearner
static void	addInterpolatedFunction (double pOut, size_t nOutVals, double pIn, size_t nInVals)
	Adds the function pIn to pOut after interpolating pIn to be the same size as pOut. (This is a helper-function used by precisionRecall.) More...

Member Enumeration Documentation

enum GClasses::GKNN::InterpolationMethod

Enumerator
Linear
Mean
Learner

enum GClasses::GKNN::TrainMethod

Enumerator
StoreAll
ValidationPrune
DrawRandom

Constructor & Destructor Documentation

GClasses::GKNN::GKNN ( )

General-purpose constructor.

GClasses::GKNN::GKNN ( const GDomNode * pNode )

Load from a DOM.

virtual GClasses::GKNN::~GKNN ( )

virtual

Member Function Documentation

size_t GClasses::GKNN::addVector	(	const GVec &	in,
		const GVec &	out
	)

Adds a copy of pVector to the internal set.

void GClasses::GKNN::autoTune	(	GMatrix &	features,
		GMatrix &	labels
	)

Uses cross-validation to find a set of parameters that works well with the provided data.

virtual void GClasses::GKNN::beginIncrementalLearningInner	(	const GRelation &	featureRel,
		const GRelation &	labelRel
	)

protectedvirtual

See the comment for GIncrementalLearner::beginIncrementalLearningInner.

Implements GClasses::GIncrementalLearner.

virtual bool GClasses::GKNN::canImplicitlyHandleMissingFeatures ( )

inlineprotectedvirtual

See the comment for GTransducer::canImplicitlyHandleMissingFeatures.

Reimplemented from GClasses::GTransducer.

virtual void GClasses::GKNN::clear ( )

virtual

Discard any training (but not any settings) so it can be trained again.

Implements GClasses::GSupervisedLearner.

void GClasses::GKNN::drawRandom ( size_t n )

inline

Specify to train by drawing 'n' random patterns from the training set.

GMatrix* GClasses::GKNN::features ( )

inline

Returns the internal feature set.

size_t GClasses::GKNN::findNeighbors ( const GVec & vector )

protected

Finds the nearest neighbors of pVector. Returns the number of neighbors found.

void GClasses::GKNN::interpolateLearner	(	size_t	nc,
		const GVec &	in,
		GPrediction *	pOut,
		GVec *	pOut2
	)

protected

Interpolates with the provided supervised learning algorithm.

void GClasses::GKNN::interpolateLinear	(	size_t	nc,
		const GVec &	in,
		GPrediction *	pOut,
		GVec *	pOut2
	)

protected

Interpolate with each neighbor having a linear vote. (Actually it's linear with respect to the squared distance instead of the distance, because this is faster to compute.)

void GClasses::GKNN::interpolateMean	(	size_t	nc,
		const GVec &	in,
		GPrediction *	pOut,
		GVec *	pOut2
	)

protected

Interpolate with each neighbor having equal vote.

GMatrix* GClasses::GKNN::labels ( )

inline

Returns the internal label set.

GDistanceMetric* GClasses::GKNN::metric ( )

inline

Returns the dissimilarity metric.

size_t GClasses::GKNN::neighborCount ( )

inline

Returns the number of neighbors.

virtual void GClasses::GKNN::predict	(	const GVec &	in,
		GVec &	out
	)

virtual

See the comment for GSupervisedLearner::predict.

Implements GClasses::GSupervisedLearner.

virtual void GClasses::GKNN::predictDistribution	(	const GVec &	in,
		GPrediction *	pOut
	)

virtual

See the comment for GSupervisedLearner::predictDistribution.

Implements GClasses::GSupervisedLearner.

virtual GDomNode* GClasses::GKNN::serialize ( GDom * pDoc ) const

virtual

Marshal this object into a DOM, which can then be converted to a variety of serial formats.

Implements GClasses::GSupervisedLearner.

void GClasses::GKNN::setInterpolationLearner	(	GSupervisedLearner *	pLearner,
		bool	bTakeOwnership
	)

Sets the interpolation method to "Learner" and sets the learner to use. If bTakeOwnership is true, it will delete the learner when this object is deleted.

void GClasses::GKNN::setInterpolationMethod ( InterpolationMethod eMethod )

Sets the technique for interpolation. (If you want to use the "Learner" method, you should call SetInterpolationLearner instead of this method.)

void GClasses::GKNN::setMetric	(	GDistanceMetric *	pMetric,
		bool	own
	)

Sets the distance metric to use for finding neighbors. If own is true, then this object will delete pMetric when it is done with it.

void GClasses::GKNN::setMetric	(	GSparseSimilarity *	pMetric,
		bool	own
	)

Sets the sparse similarity metric to use for finding neighbors. If own is true, then this object will delete pMetric when it is done with it.

void GClasses::GKNN::setNeighborCount ( size_t k )

Specify the number of neighbors to use. (The default is 1.)

void GClasses::GKNN::setNormalizeScaleFactors ( bool b )

Specify whether to normalize the scaling of each attribute. (The default is to normalize.)

void GClasses::GKNN::setOptimizeScaleFactors ( bool b )

If you set this to true, it will use a hill-climber to optimize the attribute scaling factors. If you set it to false (the default), it won't.

GSparseMatrix* GClasses::GKNN::sparseFeatures ( )

inline

Returns the internal set of sparse features.

static void GClasses::GKNN::test ( )

static

Performs unit tests for this class. Throws an exception if there is a failure.

virtual void GClasses::GKNN::trainIncremental	(	const GVec &	in,
		const GVec &	out
	)

protectedvirtual

Adds a vector to the internal set. Also, if the (k+1)th nearest neighbor of that vector is less than "elbow room" from it, then the closest neighbor is deleted from the internal set. (You might be wondering why the decision to delete the closest neighbor is determined by the distance of the (k+1)th neigbor. This enables a clump of k points to form in the most frequently sampled locations. Also, If you make this decision based on a closer neighbor, then big holes may form in the model if points are sampled in a poor order.) Call SetElbowRoom to specify the elbow room distance.

Implements GClasses::GIncrementalLearner.

virtual void GClasses::GKNN::trainInner	(	const GMatrix &	features,
		const GMatrix &	labels
	)

protectedvirtual

See the comment for GSupervisedLearner::trainInner.

Implements GClasses::GSupervisedLearner.

virtual void GClasses::GKNN::trainSparse	(	GSparseMatrix &	features,
		GMatrix &	labels
	)

virtual

See the comment for GIncrementalLearner::trainSparse.

Implements GClasses::GIncrementalLearner.

Member Data Documentation

bool GClasses::GKNN::m_bOwnLearner

protected

InterpolationMethod GClasses::GKNN::m_eInterpolationMethod

protected

TrainMethod GClasses::GKNN::m_eTrainMethod

protected

size_t GClasses::GKNN::m_nNeighbors

protected

bool GClasses::GKNN::m_normalizeScaleFactors

protected

bool GClasses::GKNN::m_optimizeScaleFactors

protected

bool GClasses::GKNN::m_ownMetric

protected

GKnnScaleFactorCritic* GClasses::GKNN::m_pCritic

protected

GDistanceMetric* GClasses::GKNN::m_pDistanceMetric

protected

GMatrix* GClasses::GKNN::m_pFeatures

protected

GMatrix* GClasses::GKNN::m_pLabels

protected

GSupervisedLearner* GClasses::GKNN::m_pLearner

protected

GNeighborFinderGeneralizing* GClasses::GKNN::m_pNeighborFinder

protected

GOptimizer* GClasses::GKNN::m_pScaleFactorOptimizer

protected

GSparseMatrix* GClasses::GKNN::m_pSparseFeatures

protected

GSparseSimilarity* GClasses::GKNN::m_pSparseMetric

protected

double GClasses::GKNN::m_trainParam

protected

GVec GClasses::GKNN::m_valueCounts

protected

Detailed Description

Public Types

Public Member Functions

Static Public Member Functions

Protected Member Functions

Protected Attributes

Additional Inherited Members

Member Enumeration Documentation

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation