classifiers.vdm
Class ValueDifferenceMetric

java.lang.Object
  |
  +--coreComponents.DistanceMetric
        |
        +--classifiers.vdm.ValueDifferenceMetric

All Implemented Interfaces:

java.lang.Cloneable, NonExchangeableDistance, java.io.Serializable

public class ValueDifferenceMetric

extends DistanceMetric

implements NonExchangeableDistance, java.lang.Cloneable, java.io.Serializable

Implementation of the Value Difference Metric used in the PEBLS nearest neighbour algorithm detailed in the paper: A Weighted Nearest Neighbor Algorithm for Learning with Symbolic Features 1993, S. Cost and S. Salzberg, John Hopkins University, Baltimore

Version:

$Version: 1.0$

Author:

David Lindsay (davidl@cs.rhul.ac.uk)

See Also:

Serialized Form

Field Summary

Instances	m_DistanceTrain The training set used to generate the VD Matrices
double	m_K The type of Minkowski type distance to use (default K=1 Manhatten)
ValueDifferenceMetric	m_TentativeDistanceMetric
classifiers.vdm.ValueDifferenceMetric.VDMatrix[]	m_VDMatrix The VD Matrices for each attribute

Constructor Summary

ValueDifferenceMetric()
Default constructor

ValueDifferenceMetric(Instances insts)
More complicated constructor

Method Summary

double	calculateTentativeDistance(Instance inst1, Instance inst2) Calculate the tentative instance of the classifier.
boolean	checkInstances(Instances insts) Check if the instances are valid for the distance metric.
double	computeValueDifference(int attIndex, int attValue1, int attValue2)
double	distance(Instance inst1, Instance inst2) Calculates the distance between two instances.
static void	main(java.lang.String[] args)
void	reset(Instances insts) This function is useful if we need to reinitialise the distance metric without using the constructor
boolean	setTentativeInstance(Instance inst) Sets the instances which we tentatively update the distance metric for.
java.lang.String	toString() Creates a debugging string detailing the information about the distance metric
void	updateRanges(Instance inst) Update the ranges of the distance metric with a new training instance.

Methods inherited from class coreComponents.DistanceMetric

forName

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, wait, wait, wait

Field Detail

m_VDMatrix

public classifiers.vdm.ValueDifferenceMetric.VDMatrix[] m_VDMatrix

The VD Matrices for each attribute

m_DistanceTrain

public Instances m_DistanceTrain

The training set used to generate the VD Matrices

m_K

public double m_K

The type of Minkowski type distance to use (default K=1 Manhatten)

m_TentativeDistanceMetric

public ValueDifferenceMetric m_TentativeDistanceMetric

Constructor Detail

ValueDifferenceMetric

public ValueDifferenceMetric()

Default constructor

ValueDifferenceMetric

public ValueDifferenceMetric(Instances insts)

More complicated constructor

Method Detail

setTentativeInstance

public boolean setTentativeInstance(Instance inst)

Sets the instances which we tentatively update the distance metric for.

Specified by:

setTentativeInstance in interface NonExchangeableDistance

calculateTentativeDistance

public double calculateTentativeDistance(Instance inst1,
                                         Instance inst2)

Calculate the tentative instance of the classifier.

Specified by:

calculateTentativeDistance in interface NonExchangeableDistance

reset

public void reset(Instances insts)

This function is useful if we need to reinitialise the distance metric without using the constructor

Specified by:

reset in class DistanceMetric

updateRanges

public void updateRanges(Instance inst)

Description copied from class: DistanceMetric

Update the ranges of the distance metric with a new training instance.

Specified by:

updateRanges in class DistanceMetric

Parameters:

inst - the instance to update with

Returns:

nothing

computeValueDifference

public double computeValueDifference(int attIndex,
                                     int attValue1,
                                     int attValue2)

distance

public double distance(Instance inst1,
                       Instance inst2)

Description copied from class: DistanceMetric

Calculates the distance between two instances.

Specified by:

distance in class DistanceMetric

Parameters:

inst1 - the first instance

inst2 - the second instance

Returns:

the distance between the two instances

toString

public java.lang.String toString()

Description copied from class: DistanceMetric

Creates a debugging string detailing the information about the distance metric

Specified by:

toString in class DistanceMetric

Returns:

a debugging string

checkInstances

public boolean checkInstances(Instances insts)

Description copied from class: DistanceMetric

Check if the instances are valid for the distance metric.

Specified by:

checkInstances in class DistanceMetric

Parameters:

insts - the training instances to be checked

Returns:

true if the instances are valid, false otherwise

main

public static void main(java.lang.String[] args)