jsat.distributions.empirical

Class KernelDensityEstimator

java.lang.Object

jsat.distributions.Distribution

jsat.distributions.ContinuousDistribution

jsat.distributions.empirical.KernelDensityEstimator

All Implemented Interfaces:

Serializable, Cloneable

public class KernelDensityEstimator
extends ContinuousDistribution

Kernel Density Estimator, KDE, uses the data set itself to approximate the underlying probability distribution using Kernel Functions.

Author:

Edward Raff

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
KernelDensityEstimator(Vec dataPoints)
KernelDensityEstimator(Vec dataPoints, KernelFunction k)
KernelDensityEstimator(Vec dataPoints, KernelFunction k, double h)
KernelDensityEstimator(Vec dataPoints, KernelFunction k, double[] weights)
KernelDensityEstimator(Vec dataPoints, KernelFunction k, double h, double[] weights)

Method Summary

Modifier and Type	Method and Description
static KernelFunction	autoKernel(Vec dataPoints) Automatically selects a good Kernel function for the data set that balances Execution time and accuracy
static double	BandwithGuassEstimate(Vec X)
double	cdf(double x) Computes the value of the Cumulative Density Function (CDF) at the given point.
KernelDensityEstimator	clone()
boolean	equals(Object obj)
double	getBandwith()
double[]	getCurrentVariableValues() Returns an array, where each value contains the value of a parameter in the distribution.
String	getDistributionName() Return the name of the distribution.
String[]	getVariables() Returns an array, where each value contains the name of a parameter in the distribution.
int	hashCode()
double	invCdf(double p) Computes the inverse Cumulative Density Function (CDF-1) at the given point.
double	max() The maximum value for which the #pdf(double) is meant to return a value.
double	mean() Computes the mean value of the distribution
double	min() The minimum value for which the #pdf(double) is meant to return a value.
double	mode() Computes the mode of the distribution.
double	pdf(double x) Computes the value of the Probability Density Function (PDF) at the given point
void	setBandwith(double val) Sets the bandwidth used for smoothing.
void	setUsingData(Vec data) Attempts to set the variables used by this distribution based on population sample data, assuming the sample data is from this type of distribution.
void	setVariable(String var, double value) Sets one of the variables of this distribution by the name.
double	skewness() Computes the skewness of the distribution.
double	variance() Computes the variance of the distribution.

Methods inherited from class jsat.distributions.ContinuousDistribution

getDescriptiveName, getFunctionPDF, getIntegrationMax, getIntegrationMin, logPdf, toString

Methods inherited from class jsat.distributions.Distribution

getFunctionCDF, invCdf, median, sample, sampleVec, standardDeviation

Methods inherited from class java.lang.Object

finalize, getClass, notify, notifyAll, wait, wait, wait

Constructor Detail

KernelDensityEstimator

public KernelDensityEstimator(Vec dataPoints)

KernelDensityEstimator

public KernelDensityEstimator(Vec dataPoints,
                              KernelFunction k)

KernelDensityEstimator

public KernelDensityEstimator(Vec dataPoints,
                              KernelFunction k,
                              double[] weights)

KernelDensityEstimator

public KernelDensityEstimator(Vec dataPoints,
                              KernelFunction k,
                              double h)

KernelDensityEstimator

public KernelDensityEstimator(Vec dataPoints,
                              KernelFunction k,
                              double h,
                              double[] weights)

Method Detail

BandwithGuassEstimate

public static double BandwithGuassEstimate(Vec X)

autoKernel

public static KernelFunction autoKernel(Vec dataPoints)

Automatically selects a good Kernel function for the data set that balances Execution time and accuracy

Parameters:

dataPoints -

Returns:

a kernel that will work well for the given distribution

pdf

public double pdf(double x)

Description copied from class: ContinuousDistribution

Computes the value of the Probability Density Function (PDF) at the given point

Specified by:

pdf in class ContinuousDistribution

Parameters:

x - the value to get the PDF

Returns:

the PDF(x)

cdf

public double cdf(double x)

Description copied from class: Distribution

Computes the value of the Cumulative Density Function (CDF) at the given point. The CDF returns a value in the range [0, 1], indicating what portion of values occur at or below that point.

Overrides:

cdf in class ContinuousDistribution

Parameters:

x - the value to get the CDF of

Returns:

the CDF(x)

invCdf

public double invCdf(double p)

Description copied from class: Distribution

Computes the inverse Cumulative Density Function (CDF^-1) at the given point. It takes in a value in the range of [0, 1] and returns the value x, such that CDF(x) = p

Overrides:

invCdf in class ContinuousDistribution

Parameters:

p - the probability value

Returns:

the value such that the CDF would return p

min

public double min()

Description copied from class: Distribution

The minimum value for which the #pdf(double) is meant to return a value. Note that Double.NEGATIVE_INFINITY is a valid return value.

Specified by:

min in class Distribution

Returns:

the minimum value for which the #pdf(double) is meant to return a value.

max

public double max()

Description copied from class: Distribution

The maximum value for which the #pdf(double) is meant to return a value. Note that Double.POSITIVE_INFINITY is a valid return value.

Specified by:

max in class Distribution

Returns:

the maximum value for which the #pdf(double) is meant to return a value.

getDistributionName

public String getDistributionName()

Description copied from class: ContinuousDistribution

Return the name of the distribution.

Specified by:

getDistributionName in class ContinuousDistribution

Returns:

the name of the distribution.

getVariables

public String[] getVariables()

Description copied from class: ContinuousDistribution

Returns an array, where each value contains the name of a parameter in the distribution. The order must always be the same, and match up with the values returned by ContinuousDistribution.getCurrentVariableValues()

Specified by:

getVariables in class ContinuousDistribution

Returns:

a string of the variable names this distribution uses

getCurrentVariableValues

public double[] getCurrentVariableValues()

Description copied from class: ContinuousDistribution

Returns an array, where each value contains the value of a parameter in the distribution. The order must always be the same, and match up with the values returned by ContinuousDistribution.getVariables()

Specified by:

getCurrentVariableValues in class ContinuousDistribution

Returns:

the current values of the parameters used by this distribution, in the same order as their names are returned by ContinuousDistribution.getVariables()

setBandwith

public void setBandwith(double val)

Sets the bandwidth used for smoothing. Higher values make the pdf smoother, but can obscure features. Too small a bandwidth will causes spikes at only the data points.

Parameters:

val - new bandwidth

getBandwith

public double getBandwith()

Returns:

the bandwidth parameter

setVariable

public void setVariable(String var,
                        double value)

Description copied from class: ContinuousDistribution

Sets one of the variables of this distribution by the name.

Specified by:

setVariable in class ContinuousDistribution

Parameters:

var - the variable to set

value - the value to set

clone

public KernelDensityEstimator clone()

Specified by:

clone in class ContinuousDistribution

setUsingData

public void setUsingData(Vec data)

Description copied from class: ContinuousDistribution

Attempts to set the variables used by this distribution based on population sample data, assuming the sample data is from this type of distribution.

Specified by:

setUsingData in class ContinuousDistribution

Parameters:

data - the data to use to attempt to fit against

mean

public double mean()

Description copied from class: Distribution

Computes the mean value of the distribution

Overrides:

mean in class ContinuousDistribution

Returns:

the mean value of the distribution

mode

public double mode()

Description copied from class: Distribution

Computes the mode of the distribution. Not all distributions have a mode for all parameter values. NaN may be returned if the mode is not defined for the current values of the distribution.

Overrides:

mode in class ContinuousDistribution

Returns:

the mode of the distribution

variance

public double variance()

Description copied from class: Distribution

Computes the variance of the distribution. Not all distributions have a finite variance for all parameter values. NaN may be returned if the variance is not defined for the current values of the distribution. Infinity is a possible value to be returned by some distributions.

Overrides:

variance in class ContinuousDistribution

Returns:

the variance of the distribution.

skewness

public double skewness()

Description copied from class: Distribution

Computes the skewness of the distribution. Not all distributions have a finite skewness for all parameter values. NaN may be returned if the skewness is not defined for the current values of the distribution.

Overrides:

skewness in class ContinuousDistribution

Returns:

the skewness of the distribution.

hashCode

public int hashCode()

Overrides:

hashCode in class Object

equals

public boolean equals(Object obj)

Overrides:

equals in class Object