jphase.fit
Class EMHyperErlangFit

java.lang.Object
  jphase.fit.ContPhaseFitter
      jphase.fit.MLContPhaseFitter
          jphase.fit.EMHyperErlangFit

All Implemented Interfaces:

PhaseFitter

public class EMHyperErlangFit
extends MLContPhaseFitter

This class implements the Maximum Likelihood method proposed by Th�mmler, Buchholz and Telek in "A novel approach for fitting probability distributions to real trace data with the EM algorithm", 2005. The method matches the likelilihood of any distribution to a subclass of Phase-Type distributions known as Hyper-Erlang distributions.

Author:

Juan Fernando Perez

Field Summary

static double	precision Precision for the convergence criterion in the algorithm
static double	precisionCV Precision for the convergence criterion in the coefficient of variance

Fields inherited from class jphase.fit.ContPhaseFitter

data, var

Constructor Summary

EMHyperErlangFit(double[] data)

Method Summary

HyperErlangVar	doFitHyperErlang(double[] data) Returns a HyperErlang variable with the best fit
double	doFitNM(double[] data, HyperErlangVar var) This method returns a completely specified HyperErlang variable, such that it has the best likelihood between all the possible combinations of N phases in M branches
double	doFitNMR(double[] data, HyperErlangVar var) This method returns a completely specified HyperErlang variable, such that it has the best likelihood after the execution of the EM algorithm for the case where the variable has N phases in M branches, distributed as determined by the vector r
DenseContPhaseVar	fit() Returns a HyperErlang variable with the best fit, in the form of a Dense Continuous Phase variable

Methods inherited from class jphase.fit.MLContPhaseFitter

getLogLikelihood

Methods inherited from class java.lang.Object

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

Field Detail

precision

public static double precision

Precision for the convergence criterion in the algorithm

precisionCV

public static double precisionCV

Precision for the convergence criterion in the coefficient of variance

Constructor Detail

EMHyperErlangFit

public EMHyperErlangFit(double[] data)

Parameters:

data -

Method Detail

fit

public DenseContPhaseVar fit()

Returns a HyperErlang variable with the best fit, in the form of a Dense Continuous Phase variable

Specified by:

fit in interface PhaseFitter

Specified by:

fit in class ContPhaseFitter

Returns:

HyperErlang variable with the best fit

See Also:

PhaseFitter.fit()

doFitHyperErlang

public HyperErlangVar doFitHyperErlang(double[] data)

Returns a HyperErlang variable with the best fit

Parameters:

data - non-negative data trace from independent experiments to be fitted

Returns:

HyperErlang variable with the best fit

doFitNM

public double doFitNM(double[] data,
                      HyperErlangVar var)

This method returns a completely specified HyperErlang variable, such that it has the best likelihood between all the possible combinations of N phases in M branches

Parameters:

data - non-negative data trace from independent experiments to be fitted

var - HyperErlang variable with the parameters N and M determined

Returns:

Likelihood of the best variable found. The variable is modified with the best parameters found.

doFitNMR

public double doFitNMR(double[] data,
                       HyperErlangVar var)

This method returns a completely specified HyperErlang variable, such that it has the best likelihood after the execution of the EM algorithm for the case where the variable has N phases in M branches, distributed as determined by the vector r

Parameters:

data - non-negative data trace from independent experiments to be fitted

var - HyperErlang variable with the parameters N, M and r determined

Returns:

Likelihood of the best variable found. The variable is modified with the best parameters found.