net.sourceforge.javaocr.plugin.cluster
Class MahalanobisDistanceCluster

java.lang.Object
  net.sourceforge.javaocr.plugin.cluster.AbstractBaseCluster
      net.sourceforge.javaocr.plugin.cluster.MahalanobisDistanceCluster

All Implemented Interfaces:

Metric, Cluster

public class MahalanobisDistanceCluster
extends AbstractBaseCluster

cluster providing Mahalanobis distance meassure ( do not ask me to pronounce this )

Author:

Konstantin Pribluda

Field Summary

(package private) double[][]	invcov
(package private) double[][]	sumxy

Constructor Summary

MahalanobisDistanceCluster()
default constructor for sake of serialisation frameworks

MahalanobisDistanceCluster(double[] mx, double[][] invcov)
convenience constructor to instantiate trained distance cluster

MahalanobisDistanceCluster(int dimensions)
constructs mahalanobis distance cluster

Method Summary

double	distance(double[] features) calculate mahalanubis distance
double[][]	getInvcov()
(package private) double[][]	matrix() calculate covariance matrix and invert it
void	setInvcov(double[][] invcov)
void	train(double[] samples) gather sampler - sum of x*y into matrix

Methods inherited from class net.sourceforge.javaocr.plugin.cluster.AbstractBaseCluster

center, getAmountSamples, getDimensions, getMx, getSum, radius, setAmountSamples, setDimensions, setMx, setSum

Methods inherited from class java.lang.Object

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

Field Detail

sumxy

double[][] sumxy

invcov

double[][] invcov

Constructor Detail

MahalanobisDistanceCluster

public MahalanobisDistanceCluster()

default constructor for sake of serialisation frameworks

MahalanobisDistanceCluster

public MahalanobisDistanceCluster(int dimensions)

constructs mahalanobis distance cluster

Parameters:

dimensions - amount of dimensions in cluster

MahalanobisDistanceCluster

public MahalanobisDistanceCluster(double[] mx,
                                  double[][] invcov)

convenience constructor to instantiate trained distance cluster

Parameters:

mx - expectation walues

invcov - inverse covariance matrix

Method Detail

distance

public double distance(double[] features)

calculate mahalanubis distance

Parameters:

features - amount of features shall correspond to amount dimensions

Returns:

calculated distance

train

public void train(double[] samples)

gather sampler - sum of x*y into matrix

Specified by:

train in interface Cluster

Overrides:

train in class AbstractBaseCluster

Parameters:

samples - sampler belonging to cluster

matrix

double[][] matrix()

calculate covariance matrix and invert it

Returns:

getInvcov

public double[][] getInvcov()

setInvcov

public void setInvcov(double[][] invcov)