org.openimaj.math.statistics.distribution

Class MultivariateKernelDensityEstimate

java.lang.Object

org.openimaj.math.statistics.distribution.AbstractMultivariateDistribution

org.openimaj.math.statistics.distribution.MultivariateKernelDensityEstimate

All Implemented Interfaces:

MultivariateDistribution

public class MultivariateKernelDensityEstimate
extends AbstractMultivariateDistribution

A Parzen window kernel density estimate using a univariate kernel and Euclidean distance. Uses a KD-Tree to for efficient neighbour search.

Author:

Jonathon Hare (jsh2@ecs.soton.ac.uk)

Constructor Summary

Constructors

Constructor and Description
MultivariateKernelDensityEstimate(double[][] data, UnivariateKernel kernel, double bandwidth) Construct with the given data, kernel and bandwidth
MultivariateKernelDensityEstimate(List<double[]> data, UnivariateKernel kernel, double bandwidth) Construct with the given data, kernel and bandwidth

Method Summary

Modifier and Type	Method and Description
double[]	estimateLogProbability(double[][] x) Get the log probability for a given point in space relative to the PDF represented by this distribution.
double	estimateProbability(double[] sample) Get the probability for a given point in space relative to the PDF represented by this distribution.
double	getBandwidth() Get the bandwidth
double[][]	getData() Get the underlying data
double	getScaledBandwidth() Get the bandwidth scaled by the kernel support.
List<ObjectDoublePair<double[]>>	getSupport(double[] sample) Get the underlying points that support the KDE within the window around the given point.
double[]	sample(Random rng) Sample the distribution.

Methods inherited from class org.openimaj.math.statistics.distribution.AbstractMultivariateDistribution

estimateLogProbability, sample

Methods inherited from class java.lang.Object

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

Constructor Detail

MultivariateKernelDensityEstimate

public MultivariateKernelDensityEstimate(double[][] data,
                                         UnivariateKernel kernel,
                                         double bandwidth)

Construct with the given data, kernel and bandwidth

Parameters:

data - the data

kernel - the kernel

bandwidth - the bandwidth

MultivariateKernelDensityEstimate

public MultivariateKernelDensityEstimate(List<double[]> data,
                                         UnivariateKernel kernel,
                                         double bandwidth)

Construct with the given data, kernel and bandwidth

Parameters:

data - the data

kernel - the kernel

bandwidth - the bandwidth

Method Detail

sample

public double[] sample(Random rng)

Description copied from interface: MultivariateDistribution

Sample the distribution. the number of samples to draw

Parameters:

rng - the random number generator

Returns:

a list of sample vectors from this distribution

estimateProbability

public double estimateProbability(double[] sample)

Description copied from interface: MultivariateDistribution

Get the probability for a given point in space relative to the PDF represented by this distribution.

Parameters:

sample - the point

Returns:

the probability

getSupport

public List<ObjectDoublePair<double[]>> getSupport(double[] sample)

Get the underlying points that support the KDE within the window around the given point. Each point is returned together with its own density estimate.

Parameters:

sample - the point in the centre of the window

Returns:

the points in the window

getData

public double[][] getData()

Get the underlying data

Returns:

the data

getBandwidth

public double getBandwidth()

Get the bandwidth

Returns:

the bandwidth

getScaledBandwidth

public double getScaledBandwidth()

Get the bandwidth scaled by the kernel support.

Returns:

the scaled bandwidth

See Also:

UnivariateKernel.getCutOff()

estimateLogProbability

public double[] estimateLogProbability(double[][] x)

Description copied from interface: MultivariateDistribution

Get the log probability for a given point in space relative to the PDF represented by this distribution.

Parameters:

x - the samples

Returns:

the log-probability