org.shared.stat.ml

Class AbstractGmModel

java.lang.Object

org.shared.stat.ml.AbstractGmModel

Direct Known Subclasses:

GmModel

public abstract class AbstractGmModel
extends Object

An abstract base class for Gaussian mixture models characterized by their initialization and update rules.

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	AbstractGmModel() Default constructor.

Method Summary

Methods

Modifier and Type	Method and Description
protected abstract RealArray	computeLogWeightedDensities(GmComponents gmc, RealArray points) Computes the log of the weighted densities.
RealArray	computePosterior(GmComponents gmc, RealArray points) Computes the posterior distribution and updates the mean likelihood of the points.
protected abstract GmComponents	initialize(RealArray points, int nComps, double regularization) Initializes the mixture components.
GmComponents	train(RealArray points, int nComps, double regularization, double delta) Trains the model.
protected abstract void	update(GmComponents gmc, RealArray posterior, RealArray points, double regularization) Performs the update rule.

Methods inherited from class java.lang.Object

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

Constructor Detail

AbstractGmModel

protected AbstractGmModel()

Default constructor.

Method Detail

train

public GmComponents train(RealArray points,
                 int nComps,
                 double regularization,
                 double delta)

Trains the model.

Parameters:

points - the input.

nComps - the number of components.

regularization - the regularization hint.

delta - the change in likelihood required for termination.

Returns:

the mixture components trained.

computePosterior

public RealArray computePosterior(GmComponents gmc,
                         RealArray points)

Computes the posterior distribution and updates the mean likelihood of the points.

Parameters:

gmc - the mixture components.

points - the input.

Returns:

the posterior distribution p(components | points).

update

protected abstract void update(GmComponents gmc,
          RealArray posterior,
          RealArray points,
          double regularization)

Performs the update rule.

Parameters:

gmc - the mixture components.

posterior - the posterior distribution p(components | points).

points - the input.

regularization - the regularization hint.

initialize

protected abstract GmComponents initialize(RealArray points,
                      int nComps,
                      double regularization)

Initializes the mixture components.

Parameters:

points - the input.

nComps - the number of components.

regularization - the regularization hint.

Returns:

the initial mixture components.

computeLogWeightedDensities

protected abstract RealArray computeLogWeightedDensities(GmComponents gmc,
                                    RealArray points)

Computes the log of the weighted densities.

Parameters:

gmc - the mixture components.

points - the input.

Returns:

the log of p(points | components).