Facemorph.Stats
Interface DataReducer

All Superinterfaces:

IniHandler, StatisticalModel

All Known Implementing Classes:

ICA, ImagePLS, OvercompleteSPLS, PCAReducer, PLSReducer, sPLSSVD, sPLSSVD_DWH, WaveletPLS, WaveletRGBPLS, WeightedPLS, WeightedPLS2

public interface DataReducer
extends StatisticalModel

This class maps a higher dimensional vector to a lower dimensional vector (and vice-versa) @Note there is not requirement for the mapping to be exactly invertable i.e. M.M^-1 != I. Only an appropriate approximation of the higher dimension is required

Nested Class Summary

static class

DataReducer.DataReducerInformation Contains information about the implementation of the DataReducer class that can be used to determine the classes abilities at run-time.

Method Summary

double	getEpsilon() Get epsilon, the very small error margin used the determine if the function has finished minimising.
double	getTargetDependentVariance()
double	getTargetIndpendentVariance()
BigMat	reconstructDependent(BigMat in) Reconstruct (approximately) a high dimensional input given the low dimensional output
double[]	reconstructDependent(double[] in)
BigMat	reconstructIndependent(BigMat in) Reconstruct (approximately) a high dimensional input given the low dimensional output
double[]	reconstructIndependent(double[] in) Reconstruct (approximately) a high dimensional input given the low dimensional output
BigMat	reduceDependent(BigMat in) Perform dimensionality reduction on the BigMat
double[]	reduceDependent(double[] in)
BigMat	reduceIndependent(BigMat in) Perform dimensionality reduction on the BigMat
double[]	reduceIndependent(double[] in) Perform dimensionality reduction on a single sample of variables
void	setEpsilon(double e) Set epsilon, the very small error margin used the determine if the function has finished minimising.
void	setOutputDimensions(int d) Set the maximum number of output dimensions in the model (pre-build, undefined if set after building some classes will alter the dimensionality of the reduction others will not) If d is greater than the number input variables, the model will truncate at the maximum number of variables.
void	setTargetDependentVariance(double var) Stop calculating components when the variance explained in the dependent variable is greater than var
void	setTargetIndependentVariance(double var) Stop calculating components when the variance explained in the independent variable is greater than var
void	setTargetKaiserGuttman()

Methods inherited from interface Facemorph.Stats.StatisticalModel

apply, getInputDimensions, getModelInformation, getModelInformation, getOutputDimensions, outputStatistics, outputStatistics, setDependentData, setDependentData, setIndependentData, train

Methods inherited from interface Facemorph.DataBase.IniHandler

read, write

Method Detail

reduceIndependent

double[] reduceIndependent(double[] in)

Perform dimensionality reduction on a single sample of variables

Parameters:

in - vector to reduce

Returns:

a shorted vector mapped in dimensionally reduced space

reconstructIndependent

double[] reconstructIndependent(double[] in)

Reconstruct (approximately) a high dimensional input given the low dimensional output

Parameters:

in - low dimensional vector

Returns:

high dimensional matrix

reduceDependent

double[] reduceDependent(double[] in)

reconstructDependent

double[] reconstructDependent(double[] in)

reduceIndependent

BigMat reduceIndependent(BigMat in)

Perform dimensionality reduction on the BigMat

Parameters:

in - BigMat to reduce

Returns:

a shorted vector mapped in dimensionally reduced space

reconstructIndependent

BigMat reconstructIndependent(BigMat in)

Reconstruct (approximately) a high dimensional input given the low dimensional output

Parameters:

in - low dimensional matrix

Returns:

high dimensional matrix

reduceDependent

BigMat reduceDependent(BigMat in)

Perform dimensionality reduction on the BigMat

Parameters:

in - BigMat to reduce

Returns:

a shorted vector mapped in dimensionally reduced space

reconstructDependent

BigMat reconstructDependent(BigMat in)

Reconstruct (approximately) a high dimensional input given the low dimensional output

Parameters:

in - low dimensional matrix

Returns:

high dimensional matrix

setOutputDimensions

void setOutputDimensions(int d)

Set the maximum number of output dimensions in the model (pre-build, undefined if set after building some classes will alter the dimensionality of the reduction others will not) If d is greater than the number input variables, the model will truncate at the maximum number of variables.

Parameters:

d - target number of dimensions.

setTargetDependentVariance

void setTargetDependentVariance(double var)

Stop calculating components when the variance explained in the dependent variable is greater than var

Parameters:

var - variance to explain

getTargetDependentVariance

double getTargetDependentVariance()

Returns:

the target variance to explain on the dependent variable (not neccessarily the amount actually calculated

setTargetIndependentVariance

void setTargetIndependentVariance(double var)

Stop calculating components when the variance explained in the independent variable is greater than var

Parameters:

var - variance to explain

getTargetIndpendentVariance

double getTargetIndpendentVariance()

Returns:

the target variance to explain on the independent variable (not neccessarily the amount actually calculated

setTargetKaiserGuttman

void setTargetKaiserGuttman()

setEpsilon

void setEpsilon(double e)

Set epsilon, the very small error margin used the determine if the function has finished minimising. i.e. if the change is less than e stop minimising

Parameters:

e - error value

getEpsilon

double getEpsilon()

Get epsilon, the very small error margin used the determine if the function has finished minimising. i.e. if the change is less than e stop minimising

Returns: