org.apache.spark.mllib.clustering

Class GaussianMixture

java.lang.Object

org.apache.spark.mllib.clustering.GaussianMixture

All Implemented Interfaces:

java.io.Serializable

public class GaussianMixture
extends java.lang.Object
implements scala.Serializable

This class performs expectation maximization for multivariate Gaussian Mixture Models (GMMs). A GMM represents a composite distribution of independent Gaussian distributions with associated "mixing" weights specifying each's contribution to the composite.

Given a set of sample points, this class will maximize the log-likelihood for a mixture of k Gaussians, iterating until the log-likelihood changes by less than convergenceTol, or until it has reached the max number of iterations. While this process is generally guaranteed to converge, it is not guaranteed to find a global optimum.

Note: For high-dimensional data (with many features), this algorithm may perform poorly. This is due to high-dimensional data (a) making it difficult to cluster at all (based on statistical/theoretical arguments) and (b) numerical issues with Gaussian distributions.

param: k Number of independent Gaussians in the mixture model. param: convergenceTol Maximum change in log-likelihood at which convergence is considered to have occurred. param: maxIterations Maximum number of iterations allowed.

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
GaussianMixture() Constructs a default instance.

Method Summary

Methods

Modifier and Type	Method and Description
double	getConvergenceTol() Return the largest change in log-likelihood at which convergence is considered to have occurred.
scala.Option<GaussianMixtureModel>	getInitialModel() Return the user supplied initial GMM, if supplied
int	getK() Return the number of Gaussians in the mixture model
int	getMaxIterations() Return the maximum number of iterations allowed
long	getSeed() Return the random seed
GaussianMixtureModel	run(JavaRDD<Vector> data) Java-friendly version of run()
GaussianMixtureModel	run(RDD<Vector> data) Perform expectation maximization
GaussianMixture	setConvergenceTol(double convergenceTol) Set the largest change in log-likelihood at which convergence is considered to have occurred.
GaussianMixture	setInitialModel(GaussianMixtureModel model) Set the initial GMM starting point, bypassing the random initialization.
GaussianMixture	setK(int k) Set the number of Gaussians in the mixture model.
GaussianMixture	setMaxIterations(int maxIterations) Set the maximum number of iterations allowed.
GaussianMixture	setSeed(long seed) Set the random seed
static boolean	shouldDistributeGaussians(int k, int d) Heuristic to distribute the computation of the MultivariateGaussians, approximately when d > 25 except for when k is very small.

Methods inherited from class java.lang.Object

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

Constructor Detail

GaussianMixture

public GaussianMixture()

Constructs a default instance. The default parameters are {k: 2, convergenceTol: 0.01, maxIterations: 100, seed: random}.

Method Detail

shouldDistributeGaussians

public static boolean shouldDistributeGaussians(int k,
                                int d)

Heuristic to distribute the computation of the MultivariateGaussians, approximately when d > 25 except for when k is very small.

Parameters:

k - Number of topics

d - Number of features

Returns:

(undocumented)

setInitialModel

public GaussianMixture setInitialModel(GaussianMixtureModel model)

Set the initial GMM starting point, bypassing the random initialization. You must call setK() prior to calling this method, and the condition (model.k == this.k) must be met; failure will result in an IllegalArgumentException

Parameters:

model - (undocumented)

Returns:

(undocumented)

getInitialModel

public scala.Option<GaussianMixtureModel> getInitialModel()

Return the user supplied initial GMM, if supplied

Returns:

(undocumented)

setK

public GaussianMixture setK(int k)

Set the number of Gaussians in the mixture model. Default: 2

Parameters:

k - (undocumented)

Returns:

(undocumented)

getK

public int getK()

Return the number of Gaussians in the mixture model

Returns:

(undocumented)

setMaxIterations

public GaussianMixture setMaxIterations(int maxIterations)

Set the maximum number of iterations allowed. Default: 100

Parameters:

maxIterations - (undocumented)

Returns:

(undocumented)

getMaxIterations

public int getMaxIterations()

Return the maximum number of iterations allowed

Returns:

(undocumented)

setConvergenceTol

public GaussianMixture setConvergenceTol(double convergenceTol)

Set the largest change in log-likelihood at which convergence is considered to have occurred.

Parameters:

convergenceTol - (undocumented)

Returns:

(undocumented)

getConvergenceTol

public double getConvergenceTol()

Return the largest change in log-likelihood at which convergence is considered to have occurred.

Returns:

(undocumented)

setSeed

public GaussianMixture setSeed(long seed)

Set the random seed

Parameters:

seed - (undocumented)

Returns:

(undocumented)

getSeed

public long getSeed()

Return the random seed

Returns:

(undocumented)

run

public GaussianMixtureModel run(RDD<Vector> data)

Perform expectation maximization

Parameters:

data - (undocumented)

Returns:

(undocumented)

run

public GaussianMixtureModel run(JavaRDD<Vector> data)

Java-friendly version of run()

Parameters:

data - (undocumented)

Returns:

(undocumented)