weka.core

Class WeightedReservoirSample

java.lang.Object

weka.core.WeightedReservoirSample

All Implemented Interfaces:

java.io.Serializable, weka.core.Aggregateable<WeightedReservoirSample>

public class WeightedReservoirSample
extends java.lang.Object
implements weka.core.Aggregateable<WeightedReservoirSample>, java.io.Serializable

Class implementing weighted reservoir sampling. Can also do unweighted reservoir sampling too if the supplied weights are all 1. Is based on the idea that one way of implementing reservoir sampling is to just generate a random number (between 0 and 1) for each data point and keep the n points with the highest values.

Version:

$Revision: $

Author:

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

See Also:

Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	WeightedReservoirSample.InstanceHolder Small inner class to hold an instance an its weight.
static class	WeightedReservoirSample.InstanceHolderComparator Comparator for InstanceHolder

Constructor Summary

Constructors

Constructor and Description
WeightedReservoirSample(int size, int seed) Constructor

Method Summary

Modifier and Type	Method and Description
void	add(weka.core.Instance toAdd, double weight) (Potentially) add an instance to the reservoir
WeightedReservoirSample	aggregate(WeightedReservoirSample toAggregate)
void	finalizeAggregation()
java.util.PriorityQueue<WeightedReservoirSample.InstanceHolder>	getSample() Get the sample
weka.core.Instances	getSampleAsInstances() Get the current sample as a set of unweighted instances
weka.core.Instances	getSampleAsWeightedInstances() Get the current sample as a set of weighted instances
void	reset() Reset (clear) the reservoir

Methods inherited from class java.lang.Object

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

Constructor Detail

WeightedReservoirSample

public WeightedReservoirSample(int size,
                               int seed)

Constructor

Parameters:

size - the size of the reservoir to use

seed - the seed for random number generation

Method Detail

add

public void add(weka.core.Instance toAdd,
                double weight)

(Potentially) add an instance to the reservoir

Parameters:

toAdd - the instance to add

weight - the weight of the instance (use 1 for all instances for unweighted sampling)

aggregate

public WeightedReservoirSample aggregate(WeightedReservoirSample toAggregate)
                                  throws java.lang.Exception

Specified by:

aggregate in interface weka.core.Aggregateable<WeightedReservoirSample>

Throws:

java.lang.Exception

finalizeAggregation

public void finalizeAggregation()
                         throws java.lang.Exception

Specified by:

finalizeAggregation in interface weka.core.Aggregateable<WeightedReservoirSample>

Throws:

java.lang.Exception

getSample

public java.util.PriorityQueue<WeightedReservoirSample.InstanceHolder> getSample()

Get the sample

Returns:

the priority queue that holds the sample

getSampleAsInstances

public weka.core.Instances getSampleAsInstances()
                                         throws java.lang.Exception

Get the current sample as a set of unweighted instances

Returns:

the current sample as a set of unweighted instances

Throws:

java.lang.Exception - if we haven't seen any instances yet

getSampleAsWeightedInstances

public weka.core.Instances getSampleAsWeightedInstances()
                                                 throws java.lang.Exception

Get the current sample as a set of weighted instances

Returns:

the current sample as a set of weighted instances

Throws:

java.lang.Exception - if we haven't seen any instances yet

reset

public void reset()

Reset (clear) the reservoir