jsat.clustering

Class MeanShift

java.lang.Object

jsat.clustering.ClustererBase

jsat.clustering.MeanShift

All Implemented Interfaces:

Serializable, Clusterer

public class MeanShift
extends ClustererBase

The MeanShift algorithms performs clustering on a data set by letting the data speak for itself and performing a mode search amongst the data set, returning a cluster for each discovered mode.

While not normally discussed in the context of Mean Shift, this implementation has rudimentary outlier-removal, outliers will not be included in the clustering.

The mean shift requires a MultivariateKDE to run. Contrary to use in density estimation, where the KernelFunction used has only a minor impact on the results, it is highly recommended you use the GaussKF for the MeanShift method. This is because of the large support and better behaved derivative, which adds in the avoidance of oscillating convergence.

Implementation Note: This implementation does not snap the values to a grid. This causes the prior noted oscillation in convergence.

Author:

Edward Raff

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
static int	DefaultMaxIterations The default number of getMaxIterations() is 1000
static double	DefaultScaleBandwidthFactor The default value of getScaleBandwidthFactor() is 1.0

Constructor Summary

Constructors

Constructor and Description
MeanShift() Creates a new MeanShift clustering object using a MetricKDE, the GaussKF, and the EuclideanDistance.
MeanShift(DistanceMetric dm) Creates a new MeanShift clustering object using a MetricKDE and the GaussKF.
MeanShift(MeanShift toCopy) Copy constructor
MeanShift(MultivariateKDE mkde) Creates a new MeanShift clustering object.

Method Summary

Modifier and Type	Method and Description
MeanShift	clone()
int[]	cluster(DataSet dataSet, ExecutorService threadpool, int[] designations) Performs clustering on the given data set.
int[]	cluster(DataSet dataSet, int[] designations) Performs clustering on the given data set.
int	getMaxIterations() Returns the maximum number of iterations the algorithm will go through, terminating early if convergence has not occurred.
double	getScaleBandwidthFactor() Returns the value by which the bandwidth of the MultivariateKDE will be scaled by.
void	setMaxIterations(int maxIterations) Sets the maximum number of iterations the algorithm will go through, terminating early if convergence has not occurred.
void	setScaleBandwidthFactor(double scaleBandwidthFactor) Sets the value by which the bandwidth of the MultivariateKDE will be scaled by.

Methods inherited from class jsat.clustering.ClustererBase

cluster, cluster, createClusterListFromAssignmentArray, getDatapointsFromCluster, supportsWeightedData

Methods inherited from class java.lang.Object

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

Field Detail

DefaultMaxIterations

public static final int DefaultMaxIterations

The default number of getMaxIterations() is 1000

See Also:

Constant Field Values

DefaultScaleBandwidthFactor

public static final double DefaultScaleBandwidthFactor

The default value of getScaleBandwidthFactor() is 1.0

See Also:

Constant Field Values

Constructor Detail

MeanShift

public MeanShift()

Creates a new MeanShift clustering object using a MetricKDE, the GaussKF, and the EuclideanDistance.

MeanShift

public MeanShift(DistanceMetric dm)

Creates a new MeanShift clustering object using a MetricKDE and the GaussKF.

Parameters:

dm - the distance metric to use

MeanShift

public MeanShift(MultivariateKDE mkde)

Creates a new MeanShift clustering object.
NOTE: ProductKDE does not currently support the functions needed to work with MeanShift.

Parameters:

mkde - the KDE to use in the clustering process.

MeanShift

public MeanShift(MeanShift toCopy)

Copy constructor

Parameters:

toCopy - the object to copy

Method Detail

setMaxIterations

public void setMaxIterations(int maxIterations)

Sets the maximum number of iterations the algorithm will go through, terminating early if convergence has not occurred.

Parameters:

maxIterations - the maximum number of iterations

Throws:

ArithmeticException - if a value less than 1 is given

getMaxIterations

public int getMaxIterations()

Returns the maximum number of iterations the algorithm will go through, terminating early if convergence has not occurred.

Returns:

the maximum number of iterations

setScaleBandwidthFactor

public void setScaleBandwidthFactor(double scaleBandwidthFactor)

Sets the value by which the bandwidth of the MultivariateKDE will be scaled by.

Parameters:

scaleBandwidthFactor - the value to scale bandwidth by

Throws:

ArithmeticException - if the value given is NaN or infinity

getScaleBandwidthFactor

public double getScaleBandwidthFactor()

Returns the value by which the bandwidth of the MultivariateKDE will be scaled by.

Returns:

the value to scale bandwidth by

cluster

public int[] cluster(DataSet dataSet,
                     int[] designations)

Description copied from interface: Clusterer

Performs clustering on the given data set. Parameters may be estimated by the method, or other heuristics performed.

Parameters:

dataSet - the data set to perform clustering on

designations - the array which will contain the designated values. The array will be altered and returned by the function. If null is given, a new array will be created and returned.

Returns:

an array indicating for each value indicating the cluster designation. This is the same array as designations, or a new one if the input array was null

cluster

public int[] cluster(DataSet dataSet,
                     ExecutorService threadpool,
                     int[] designations)

Description copied from interface: Clusterer

Performs clustering on the given data set. Parameters may be estimated by the method, or other heuristics performed.

Parameters:

dataSet - the data set to perform clustering on

threadpool - a source of threads to run tasks

designations - the array which will contain the designated values. The array will be altered and returned by the function. If null is given, a new array will be created and returned.

Returns:

an array indicating for each value indicating the cluster designation. This is the same array as designations, or a new one if the input array was null

clone

public MeanShift clone()

Specified by:

clone in interface Clusterer

Specified by:

clone in class ClustererBase