gmm
Class GMMEstimator

java.lang.Object
  gmm.GMMEstimator

Direct Known Subclasses:

AStar, FVHeuristic, GreedyCSS, GreedyDK, RandomEstimator

public abstract class GMMEstimator
extends java.lang.Object

This abstract class represents a estimator of the Generalized Mallows Model. It contains generic functionality that any estimator will need. Any particular estimator must inherit from this class.

Field Summary

protected boolean	GMM
protected float	logL
protected static int	maxIters
protected int	n
protected int[]	pi
protected float[][]	Q
protected long	runningTime
protected float[]	theta
protected float[]	V

Constructor Summary

GMMEstimator(boolean GMM, float[][] Q)

Method Summary

protected void	computeLogL() Must be called by a subclass only after the theta values have been set.
protected void	computeThetas() Sets the theta values.
protected void	computeVjs() Sets the V_j values.
float	getAvgDk()
float	getLogL()
float	getLogLForQ(float[][] argQ)
int[]	getPi()
long	getRunningTime()
protected float	lnPsiJPrime(int j, float thetaJ)
protected float	minimizeF(float D)
protected float	minimizeFj(int j, float Vj)
float	minValueOfFj(int j, float Vj)
void	printLearnedModel(java.io.PrintWriter out)
protected float	PsiJ(int j, float thetaJ)
abstract void	run() This method causes the estimator to run, and estimate the GMM parameters.
protected float	sumOfLnPsiJPrime(int startJ, int endJ, float thetaJ)

Methods inherited from class java.lang.Object

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

Field Detail

GMM

protected boolean GMM

Q

protected float[][] Q

n

protected int n

pi

protected int[] pi

theta

protected float[] theta

V

protected float[] V

logL

protected float logL

runningTime

protected long runningTime

maxIters

protected static int maxIters

Constructor Detail

GMMEstimator

public GMMEstimator(boolean GMM,
                    float[][] Q)

Method Detail

run

public abstract void run()
                  throws java.lang.Exception

This method causes the estimator to run, and estimate the GMM parameters. Any estimator must provide a concrete implementation of this method.

Throws:

java.lang.Exception

getPi

public int[] getPi()

Returns:

the estimated modal ranking.

getAvgDk

public float getAvgDk()

Returns:

the average Kendall distance of estimated modal ranking pi from the rankings in the sample.

getLogL

public float getLogL()

getRunningTime

public long getRunningTime()

printLearnedModel

public void printLearnedModel(java.io.PrintWriter out)

PsiJ

protected float PsiJ(int j,
                     float thetaJ)

lnPsiJPrime

protected float lnPsiJPrime(int j,
                            float thetaJ)

Parameters:

j -

thetaJ -

Returns:

the derivative of ln psi_j(thetaJ), or g_j'(thetaJ).

sumOfLnPsiJPrime

protected float sumOfLnPsiJPrime(int startJ,
                                 int endJ,
                                 float thetaJ)

minimizeFj

protected float minimizeFj(int j,
                           float Vj)

Parameters:

j -

Vj -

Returns:

the value of theta_j >= 0 that minimizes the function f_j that comes up in GMM (multiple thetas case).

minValueOfFj

public float minValueOfFj(int j,
                          float Vj)

Parameters:

j -

Vj -

Returns:

the minimum value of the function f_j that comes up in GMM.

minimizeF

protected float minimizeF(float D)

Parameters:

D -

Returns:

the value of theta >= 0 that minimizes the function f that comes up in the Mallows Model (single theta case).

computeVjs

protected void computeVjs()

Sets the V_j values. Must be called by a subclass only after it has estimated the modal ranking pi.

computeThetas

protected void computeThetas()

Sets the theta values. Must be called by a subclass only after it has set the V_j values.

computeLogL

protected void computeLogL()

Must be called by a subclass only after the theta values have been set.

getLogLForQ

public float getLogLForQ(float[][] argQ)