PLearn::LiftStatsCollector Class Reference

#include <LiftStatsCollector.h>

Inheritance diagram for PLearn::LiftStatsCollector:

[legend]Collaboration diagram for PLearn::LiftStatsCollector:

[legend]List of all members.


Public Types
typedef VecStatsCollector	inherited
Public Member Functions
	LiftStatsCollector ()
virtual void	forget ()
	clears all previously accumulated statistics
virtual void	update (const Vec &x, real weight=1.0)
	updates the statistics when seeing x The weight applies to all elements of x
virtual void	finalize ()
	This finalize override makes sure only the n_samples_to_keep samples from the matrix n_first_updates with the highest output are left.
virtual double	getStat (const string &statspec)
	In addition to the regular VecStatsCollector statistics, we understand specific lift statistics (see the .cc).
	PLEARN_DECLARE_OBJECT (LiftStatsCollector)
virtual void	build ()
	simply calls inherited::build() then build_()
virtual void	makeDeepCopyFromShallowCopy (map< const void , void > &copies)
	Transforms a shallow copy into a deep copy.
Public Attributes
real	lift_fraction
int	opposite_lift
string	output_column
int	sign_trick
int	target_column
int	verbosity
Protected Member Functions
real	computeLift ()
	Return the LIFT statistic.
real	computeLiftMax ()
	Return the LIFT_MAX statistic.
Static Protected Member Functions
void	declareOptions (OptionList &ol)
	Declares this class' options.
Protected Attributes
int	count_fin
Mat	n_first_updates
	Matrix storing the output and target of the samples with highest output, as well as all the other data retrieved since the last call to finalize.
bool	is_finalized
	Set to true after each call to finalize().
int	nstored
	Number of examples stored in the n_first_updates matrix.
int	nsamples
	Number of samples seen.
int	npos
	Number of positive examples that are not retained in the ones with the highest output (that is to say the ones in n_first_updates).
int	n_samples_to_keep
	Number of examples to keep (nsamples * lift_fraction).
int	output_column_index
	Index of the output column.
Private Member Functions
void	build_ ()
	This does the actual building.

Member Typedef Documentation

typedef VecStatsCollector PLearn::LiftStatsCollector::inherited

Reimplemented from PLearn::VecStatsCollector.
Definition at line 57 of file LiftStatsCollector.h.
Referenced by LiftStatsCollector().

Constructor & Destructor Documentation

PLearn::LiftStatsCollector::LiftStatsCollector ( )

Definition at line 53 of file LiftStatsCollector.cc.
References inherited.

Member Function Documentation

void PLearn::LiftStatsCollector::build ( ) [virtual]

simply calls inherited::build() then build_()

Reimplemented from PLearn::VecStatsCollector.
Definition at line 133 of file LiftStatsCollector.cc.
References build_().

void PLearn::LiftStatsCollector::build_ ( ) [private]

This does the actual building.

Reimplemented from PLearn::VecStatsCollector.
Definition at line 142 of file LiftStatsCollector.cc.
References output_column, and output_column_index.
Referenced by build().

real PLearn::LiftStatsCollector::computeLift ( ) [protected]

Return the LIFT statistic.

Definition at line 160 of file LiftStatsCollector.cc.
References PLearn::TMat< T >::column(), PLearn::endl(), finalize(), is_finalized, lift_fraction, n_first_updates, n_samples_to_keep, npos, nsamples, nstored, opposite_lift, output_column, sign_trick, PLearn::sum(), target_column, and verbosity.
Referenced by getStat().

real PLearn::LiftStatsCollector::computeLiftMax ( ) [protected]

Return the LIFT_MAX statistic.

Definition at line 192 of file LiftStatsCollector.cc.
References PLearn::TMat< T >::column(), finalize(), is_finalized, MIN, n_first_updates, n_samples_to_keep, npos, nsamples, and PLearn::sum().
Referenced by getStat().

void PLearn::LiftStatsCollector::declareOptions ( OptionList & ol ) [static, protected]

Declares this class' options.

Reimplemented from PLearn::VecStatsCollector.
Definition at line 98 of file LiftStatsCollector.cc.
References PLearn::declareOption(), and PLearn::OptionList.

void PLearn::LiftStatsCollector::finalize ( ) [virtual]

This finalize override makes sure only the n_samples_to_keep samples from the matrix n_first_updates with the highest output are left.

Reimplemented from PLearn::VecStatsCollector.
Definition at line 207 of file LiftStatsCollector.cc.
References count_fin, PLearn::endl(), is_finalized, lift_fraction, n_first_updates, n_samples_to_keep, npos, nsamples, nstored, PLearn::TMat< T >::resize(), PLearn::selectAndOrder(), and verbosity.
Referenced by computeLift(), and computeLiftMax().

void PLearn::LiftStatsCollector::forget ( ) [virtual]

clears all previously accumulated statistics

Reimplemented from PLearn::VecStatsCollector.
Definition at line 249 of file LiftStatsCollector.cc.
References count_fin, is_finalized, n_first_updates, npos, nsamples, nstored, and PLearn::TMat< T >::resize().

double PLearn::LiftStatsCollector::getStat ( const string & statspec ) [virtual]

In addition to the regular VecStatsCollector statistics, we understand specific lift statistics (see the .cc).

Reimplemented from PLearn::VecStatsCollector.
Definition at line 264 of file LiftStatsCollector.cc.
References computeLift(), computeLiftMax(), and PLearn::PStream::smartReadUntilNext().

void PLearn::LiftStatsCollector::makeDeepCopyFromShallowCopy ( map< const void *, void * > & copies ) [virtual]

Transforms a shallow copy into a deep copy.

Reimplemented from PLearn::VecStatsCollector.
Definition at line 282 of file LiftStatsCollector.cc.
References PLearn::deepCopyField(), and n_first_updates.

PLearn::LiftStatsCollector::PLEARN_DECLARE_OBJECT ( LiftStatsCollector )

void PLearn::LiftStatsCollector::update ( const Vec & x,

real weight = 1.0

) [virtual]

updates the statistics when seeing x The weight applies to all elements of x

Reimplemented from PLearn::VecStatsCollector.
Definition at line 291 of file LiftStatsCollector.cc.
References count_fin, PLearn::FABS(), is_finalized, PLearn::is_missing(), PLearn::TMat< T >::length(), MAX, n_first_updates, nsamples, nstored, output_column_index, PLERROR, PLWARNING, PLearn::TMat< T >::resize(), sign_trick, target_column, PLearn::Vec, and x.

Member Data Documentation

int PLearn::LiftStatsCollector::count_fin [protected]

Definition at line 63 of file LiftStatsCollector.h.
Referenced by finalize(), forget(), and update().

bool PLearn::LiftStatsCollector::is_finalized [protected]

Set to true after each call to finalize().

Definition at line 72 of file LiftStatsCollector.h.
Referenced by computeLift(), computeLiftMax(), finalize(), forget(), and update().

real PLearn::LiftStatsCollector::lift_fraction

Definition at line 96 of file LiftStatsCollector.h.
Referenced by computeLift(), and finalize().

Mat PLearn::LiftStatsCollector::n_first_updates [protected]

Matrix storing the output and target of the samples with highest output, as well as all the other data retrieved since the last call to finalize.

Definition at line 69 of file LiftStatsCollector.h.
Referenced by computeLift(), computeLiftMax(), finalize(), forget(), makeDeepCopyFromShallowCopy(), and update().

int PLearn::LiftStatsCollector::n_samples_to_keep [protected]

Number of examples to keep (nsamples * lift_fraction).

Definition at line 85 of file LiftStatsCollector.h.
Referenced by computeLift(), computeLiftMax(), and finalize().

int PLearn::LiftStatsCollector::npos [protected]

Number of positive examples that are not retained in the ones with the highest output (that is to say the ones in n_first_updates).

Definition at line 82 of file LiftStatsCollector.h.
Referenced by computeLift(), computeLiftMax(), finalize(), and forget().