org.openimaj.math.statistics.normalisation

Class ZScore

java.lang.Object

org.openimaj.math.statistics.normalisation.ZScore

All Implemented Interfaces:

Denormaliser, Normaliser, TrainableNormaliser

public class ZScore
extends Object
implements TrainableNormaliser, Denormaliser

z-score normalisation (standardisation). Upon training, the mean and variance of each dimension is computed; normalisation works by subtracting the mean and dividing by the standard deviation.

This implementation includes an optional regularisation parameter that is added to the variance before the division.

Author:

Jonathon Hare (jsh2@ecs.soton.ac.uk)

Constructor Summary

Constructors

Constructor and Description
ZScore() Construct without regularisation.
ZScore(double eps) Construct with regularisation.

Method Summary

Modifier and Type	Method and Description
double[]	denormalise(double[] vector) Deormalise the vector.
double[][]	denormalise(double[][] data) Denormalise the data.
double[]	normalise(double[] vector) Normalise the vector.
double[][]	normalise(double[][] data) Normalise the data.
void	train(double[][] data) Train the normaliser.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

ZScore

public ZScore()

Construct without regularisation.

ZScore

public ZScore(double eps)

Construct with regularisation.

Parameters:

eps - the variance normalisation regulariser (each dimension is divided by sqrt(var + eps).

Method Detail

train

public void train(double[][] data)

Description copied from interface: TrainableNormaliser

Train the normaliser. This must be called before calling the Normaliser.normalise(double[]) or Normaliser.normalise(double[][]) methods.

Specified by:

train in interface TrainableNormaliser

Parameters:

data - the data to normalise.

normalise

public double[] normalise(double[] vector)

Description copied from interface: Normaliser

Normalise the vector.

Specified by:

normalise in interface Normaliser

Parameters:

vector - the vector

Returns:

the normalised vector

normalise

public double[][] normalise(double[][] data)

Description copied from interface: Normaliser

Normalise the data.

Specified by:

normalise in interface Normaliser

Parameters:

data - the data (one vector per row)

Returns:

the normalised data

denormalise

public double[] denormalise(double[] vector)

Description copied from interface: Denormaliser

Deormalise the vector.

Specified by:

denormalise in interface Denormaliser

Parameters:

vector - the normalised vector

Returns:

the denormalised vector

denormalise

public double[][] denormalise(double[][] data)

Description copied from interface: Denormaliser

Denormalise the data.

Specified by:

denormalise in interface Denormaliser

Parameters:

data - the normalised data (one normalised vector per row)

Returns:

the denormalised data