org.apache.ctakes.ytex.kernel

Class ImputedFeatureEvaluatorImpl.JointDistribution

java.lang.Object

org.apache.ctakes.ytex.kernel.ImputedFeatureEvaluatorImpl.JointDistribution

Enclosing class:

ImputedFeatureEvaluatorImpl

public static class ImputedFeatureEvaluatorImpl.JointDistribution
extends Object

joint distribution of concept (x) and class (y). The bins for x and y are predetermined. Typical levels for x are 0/1 (absent/present) and -1/0/1 (negated/not present/affirmed).

Author:

vijay

Field Summary

Fields

Modifier and Type	Field and Description
protected double[][]	contingencyTable
protected Double	entropyX the entropy of X.
protected Double	entropyXY the entropy of X*Y.
protected SortedMap<String,SortedMap<String,Set<Long>>>	jointDistroTable A y*x table where the cells hold the instance ids.
protected Set<String>	xVals the possible values of X (e.g.
protected Set<String>	yVals the possible values of Y (e.g.

Constructor Summary

Constructors

Constructor and Description
JointDistribution(Set<String> xVals, Set<String> yVals) set up the joint distribution table.
JointDistribution(Set<String> xVals, Set<String> yVals, Map<String,Set<Long>> xMargin, Map<String,Set<Long>> yMargin, String xLeftover)

Method Summary

Modifier and Type	Method and Description
double[][]	getContingencyTable()
double	getEntropyX()
double	getEntropyXY()
double	getInfoGain()
Set<Long>	getInstances(String x, String y)
double	getMutualInformation(double entropyY)
static ImputedFeatureEvaluatorImpl.JointDistribution	merge(List<ImputedFeatureEvaluatorImpl.JointDistribution> jointDistros, Map<String,Set<Long>> yMargin, String xMerge) merge joint distributions into a single distribution.
String	toString() print out joint distribution table

Methods inherited from class java.lang.Object

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

Field Detail

contingencyTable

protected double[][] contingencyTable

entropyX

protected Double entropyX

the entropy of X. Calculated once and returned as needed.

entropyXY

protected Double entropyXY

the entropy of X*Y. Calculated once and returned as needed.

jointDistroTable

protected SortedMap<String,SortedMap<String,Set<Long>>> jointDistroTable

A y*x table where the cells hold the instance ids. We use the instance ids instead of counts so we can merge the tables.

xVals

protected Set<String> xVals

the possible values of X (e.g. concept)

yVals

protected Set<String> yVals

the possible values of Y (e.g. text)

Constructor Detail

JointDistribution

public JointDistribution(Set<String> xVals,
                         Set<String> yVals)

set up the joint distribution table.

Parameters:

xVals - the possible x values (bins)

yVals - the possible y values (bins)

JointDistribution

public JointDistribution(Set<String> xVals,
                         Set<String> yVals,
                         Map<String,Set<Long>> xMargin,
                         Map<String,Set<Long>> yMargin,
                         String xLeftover)

Method Detail

merge

public static ImputedFeatureEvaluatorImpl.JointDistribution merge(List<ImputedFeatureEvaluatorImpl.JointDistribution> jointDistros,
                                                                  Map<String,Set<Long>> yMargin,
                                                                  String xMerge)

merge joint distributions into a single distribution. For each value of Y, the cells for each X bin, except for the xMerge bin, are the intersection of all the instances in each of the corresponding bins. The xMerge bin gets everything that is leftover.

Parameters:

jointDistros - list of joint distribution tables to merge

yMargin - map of y val - instance id. this could be calculated on the fly, but we have this information already.

xMerge - the x val that contains everything that doesn't land in any of the other bins.

Returns:

getContingencyTable

public double[][] getContingencyTable()

getEntropyX

public double getEntropyX()

getEntropyXY

public double getEntropyXY()

getInfoGain

public double getInfoGain()

getInstances

public Set<Long> getInstances(String x,
                              String y)

getMutualInformation

public double getMutualInformation(double entropyY)

toString

public String toString()

print out joint distribution table

Overrides:

toString in class Object