com.penguinwerks.jodene.data
Class ClassificationScaler

java.lang.Object
  com.penguinwerks.jodene.data.Scaler
      com.penguinwerks.jodene.data.ClassificationScaler

All Implemented Interfaces:

java.io.Serializable

public class ClassificationScaler

extends Scaler

implements java.io.Serializable

This scaler is used to translate an input value for a classifier, 1.0 or 0.0 usually, to an intermediate value like 0.1 and 0.9. This can sometimes aid training because 1.0 and 0.0 sometimes drive the network towards saturation - for example trying to achieve a 1.0 response from a sigmoid is impossible.

Author:

Paul Hoehne

See Also:

Serialized Form

Constructor Summary

ClassificationScaler()
Constructor - no argument constructor.

ClassificationScaler(double min, double max)
Constructor - sets the min and the max.

Method Summary

double	getMaxScale() This is the maximum value for a classifier, usually considered the "on" or "yes" value.
double	getMinScale() This is the minimum value for a classifier, usually considered the "off" or "no" value.
double	scale(double input) Scales the input value.
void	setMaxScale(double maxScale) This is the maximum value for a classifier, usually considered the "on" or "yes" value.
void	setMinScale(double minScale) This is the minimum value for a classifier, usually considered the "off" or "no" value.

Methods inherited from class com.penguinwerks.jodene.data.Scaler

addScaler

Methods inherited from class java.lang.Object

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

Constructor Detail

ClassificationScaler

public ClassificationScaler()

Constructor - no argument constructor.

ClassificationScaler

public ClassificationScaler(double min,
                            double max)

Constructor - sets the min and the max.

Parameters:

min - The minimum end of the scale.

max - The maximum end of the scale.

Method Detail

getMaxScale

public double getMaxScale()

This is the maximum value for a classifier, usually considered the "on" or "yes" value. Normally this is 0.9.

Returns:

Returns the maxScale.

setMaxScale

public void setMaxScale(double maxScale)

This is the maximum value for a classifier, usually considered the "on" or "yes" value. Normally this is 0.9.

Parameters:

maxScale - The maxScale to set.

getMinScale

public double getMinScale()

This is the minimum value for a classifier, usually considered the "off" or "no" value. Normally this is 0.1.

Returns:

Returns the minScale.

setMinScale

public void setMinScale(double minScale)

This is the minimum value for a classifier, usually considered the "off" or "no" value. Normally this is 0.1.

Parameters:

minScale - The minScale to set.

scale

public double scale(double input)

Scales the input value. Given an input value scales it to the appropriate response. Overrides the abstract method on Scaler.

Specified by:

scale in class Scaler

Parameters:

input - The unscaled input value.

Returns:

The scaled value.