weka.core.stats

Class TDigest

java.lang.Object

weka.core.stats.TDigest

All Implemented Interfaces:

java.io.Serializable

public abstract class TDigest
extends java.lang.Object
implements java.io.Serializable

Wrapper for TDigest quantile estimators. Currently provides the clearspring analytics version and the latest and greatest from Ted Dunning.

Version:

$Revision: $

Author:

Mark Hall (mhall{[at]}pentaho{[dot]}com)

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
TDigest()

Method Summary

Modifier and Type	Method and Description
void	add(double x) Add a point to the estimator
abstract void	add(double x, int w) Add a point to the estimator
abstract void	asBytes(java.nio.ByteBuffer buff) Encode the estimator into a byte buffer
abstract void	asSmallBytes(java.nio.ByteBuffer buff) More compact encoding
abstract int	byteSize() Number of bytes required for the standard encoding
abstract double	cdf(double x) CDF
abstract double	compression() Get the current compression level
static TDigest	createTDigest(double compression) Factory method.
static TDigest	fromBytes(java.nio.ByteBuffer buff) Decode a TDigest estimator from a byte buffer
static TDigest	merge(double compression, java.lang.Iterable<TDigest> subData) Merge a collection of TDigest estimators into one
abstract double	quantile(double x) Quantile estimate
abstract int	smallByteSize() Number of bytes required for the compact encoding

Methods inherited from class java.lang.Object

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

Constructor Detail

TDigest

public TDigest()

Method Detail

createTDigest

public static TDigest createTDigest(double compression)

Factory method. Currently returns Ted's AVL tree method

Parameters:

compression - the compression level to use

Returns:

a TDigest estimator

add

public void add(double x)

Add a point to the estimator

Parameters:

x - the point to add

add

public abstract void add(double x,
                         int w)

Add a point to the estimator

Parameters:

x - the point to add

w - the weight of the point

cdf

public abstract double cdf(double x)

CDF

Parameters:

x - point at which the CDF should be evaluated

Returns:

the approximate fraction of all samples that were <= x

quantile

public abstract double quantile(double x)

Quantile estimate

Parameters:

x - which quantile to return

Returns:

the quantile estimate

asBytes

public abstract void asBytes(java.nio.ByteBuffer buff)

Encode the estimator into a byte buffer

Parameters:

buff - the buffer to hold the encoded estimator

asSmallBytes

public abstract void asSmallBytes(java.nio.ByteBuffer buff)

More compact encoding

Parameters:

buff - the buffer to hold the encoded estiamtor

byteSize

public abstract int byteSize()

Number of bytes required for the standard encoding

Returns:

the number of bytes required

smallByteSize

public abstract int smallByteSize()

Number of bytes required for the compact encoding

Returns:

the number of bytes required

compression

public abstract double compression()

Get the current compression level

Returns:

the compression level

fromBytes

public static TDigest fromBytes(java.nio.ByteBuffer buff)

Decode a TDigest estimator from a byte buffer

Parameters:

buff - the buffer

Returns:

the decoded TDigest estimator

merge

public static TDigest merge(double compression,
                            java.lang.Iterable<TDigest> subData)

Merge a collection of TDigest estimators into one

Parameters:

compression - the compression level

subData - the collection of estimators to merge

Returns:

a single merged estimator