moa.clusterers.clustree

Class Entry

java.lang.Object

moa.clusterers.clustree.Entry

public class Entry
extends Object

Field Summary

Fields

Modifier and Type	Field and Description
ClusKernel	data The actual entry data.

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	Entry(Entry other) Copy constructor.
	Entry(int numberDimensions) Constructor for the entry.
	Entry(int numberDimensions, ClusKernel cluster, long currentTime) Constructuctor that creates an Entry with an empty buffer and the data given by the Kernel.
protected	Entry(int numberDimensions, ClusKernel cluster, long currentTime, Entry parentEntry, Node containerNode) extended constructor with containerNode and parentEntry
protected	Entry(int numberDimensions, Node node, long currentTime, Entry parentEntry, Node containerNode) Constructor that creates an Entry that points to the given node.

Method Summary

Methods

Modifier and Type	Method and Description
void	add(Entry other) Add the data cluster of another entry to the data cluster of this entry.
protected void	aggregateCluster(ClusKernel otherData, long currentTime, double negLambda) Aggregate the given Kernel to the data cluster of this entry.
protected void	aggregateEntry(Entry other, long currentTime, double negLambda) Aggregate the data in the Kernel of the other Entry.
protected void	aggregateToBuffer(ClusKernel pointToInsert, long currentTime, double negLambda) Aggregate the given Kernel to the buffer cluster of this entry.
protected double	calcDistance(ClusKernel cluster) Calculates the distance to the data in this entry.
double	calcDistance(Entry other) Calculates the distance to the data in this entry of the data in the given entry.
protected void	clear() Clear the Entry.
protected ClusKernel	emptyBuffer(long currentTime, double negLambda) Clear the buffer in this entry and return a copy.
protected ClusKernel	getBuffer() Getter for the buffer.
Node	getChild() Return the reference to the child of this Entry to navigate in the tree.
protected ClusKernel	getData() Getter for the data.
Node	getNode()
Entry	getParentEntry()
long	getTimestamp() Return the current timestamp.
protected void	initializeEntry(Entry other, long currentTime) When this entry is empty, give it it's first values.
protected boolean	isEmpty() Check if this Entry is empty or not.
protected boolean	isIrrelevant(double threshold) Returns true if this entry is irrelevant with respecto the given threshold.
protected void	makeOlder(long currentTime, double negLambda) Ages this entrie's data AND buffer according to the given time and aging constant.
protected void	mergeWith(Entry other) Merge this entry witht the given Entry.
protected void	overwriteOldEntry(Entry newEntry) Overwrites the LS, SS and weightedN in the data cluster of this Entry to the values of the data cluster in the given Entry, but adds N and classCount of the cluster in the given Entry to the data cluster in this one.
protected void	recalculateData() This functions reads every entry in the child node and calculates the corresponding data Kernel.
void	setChild(Node child) Setter for the child in this entry.
void	setNode(Node node)
void	setParentEntry(Entry parent)
protected void	shallowClear() Clear the data and the buffer Custer in this entry.

Methods inherited from class java.lang.Object

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

Field Detail

data

public ClusKernel data

The actual entry data.

Constructor Detail

Entry

public Entry(int numberDimensions)

Constructor for the entry. To be used when we want to create an empty entry. Notice that the timestamp will be set to zero, since there is no reason to know when an empty entry was generated.

Parameters:

numberDimensions - The dimensionality of the data point in tree where this entry is used.

Entry

protected Entry(int numberDimensions,
     Node node,
     long currentTime,
     Entry parentEntry,
     Node containerNode)

Constructor that creates an Entry that points to the given node. The values of data will be calculated for this.

Parameters:

numberDimensions - The dimensionality of the node.

node - The node to which the new Entry should point.

currentTime - The timestamp for the moment where this Entry was was generated.

See Also:

Node, data

Entry

public Entry(int numberDimensions,
     ClusKernel cluster,
     long currentTime)

Constructuctor that creates an Entry with an empty buffer and the data given by the Kernel.

Parameters:

numberDimensions - The dimensionality of the information in the cluster.

cluster - The cluster from which the information is to be extracted.

currentTime - The timestamp for the moment where this Entry was was generated.

See Also:

Kernel, data

Entry

protected Entry(int numberDimensions,
     ClusKernel cluster,
     long currentTime,
     Entry parentEntry,
     Node containerNode)

extended constructor with containerNode and parentEntry

Parameters:

numberDimensions -

cluster -

currentTime -

parentEntry -

containerNode -

Entry

protected Entry(Entry other)

Copy constructor. Everythin is copied, including the child.

Parameters:

other -

Method Detail

getNode

public Node getNode()

setNode

public void setNode(Node node)

clear

protected void clear()

Clear the Entry. All points in the buffer and in the data cluster are lost, the connection to the child is lost and the timestamp is set to the default value.

shallowClear

protected void shallowClear()

Clear the data and the buffer Custer in this entry. This function does not clear the child of this Entry.

See Also:

data, buffer, Kernel

calcDistance

protected double calcDistance(ClusKernel cluster)

Calculates the distance to the data in this entry.

Parameters:

cluster - The Kernel cluster to which the distance is to be calculated.

Returns:

The distance to the data Kernel in this Entry

See Also:

Kernel, data

calcDistance

public double calcDistance(Entry other)

Calculates the distance to the data in this entry of the data in the given entry.

Parameters:

other - The Entry to which the distance is to be calculated.

Returns:

The distance to the data Kernel in this Entry of the data Kernel in the other Entry.

See Also:

Kernel, data

initializeEntry

protected void initializeEntry(Entry other,
                   long currentTime)

When this entry is empty, give it it's first values. It makes sense to have this operation separated from the aggregation, because the aggregation first weights the values in data and Kernel, which makes no sense in an empty entry.

Parameters:

other - The entry with the information to be used to initialize this entry.

currentTime - The time at which this is happening.

add

public void add(Entry other)

Add the data cluster of another entry to the data cluster of this entry. By using this function the timestamp does not get updated, nor does this entry get older.

Parameters:

other - The entry of which the data cluster should be added to the local data cluster.

See Also:

data, Kernel#add(tree.Kernel)

aggregateEntry

protected void aggregateEntry(Entry other,
                  long currentTime,
                  double negLambda)

Aggregate the data in the Kernel of the other Entry.

Parameters:

other - The Entry to be aggregated.

See Also:

data, Kernel

aggregateCluster

protected void aggregateCluster(ClusKernel otherData,
                    long currentTime,
                    double negLambda)

Aggregate the given Kernel to the data cluster of this entry.

Parameters:

otherData - The Entry to be aggregated.

See Also:

data, Kernel

aggregateToBuffer

protected void aggregateToBuffer(ClusKernel pointToInsert,
                     long currentTime,
                     double negLambda)

Aggregate the given Kernel to the buffer cluster of this entry.

Parameters:

pointToInsert - The cluster to aggregate to the buffer.

currentTime - The time at which the aggregation occurs.

negLambda - A parameter needed to weight the current state of the buffer.

mergeWith

protected void mergeWith(Entry other)

Merge this entry witht the given Entry. This adds the data cluster of the given Entry to the data cluster of this entry and sets the timestamp to the newest one of the the two entries.

Parameters:

other - The entry from which the data cluster is added.

See Also:

Kernel#add(tree.Kernel)

getBuffer

protected ClusKernel getBuffer()

Getter for the buffer. It is the real object, that means side effects are possible!

Returns:

A reference to the buffer in this entry.

getChild

public Node getChild()

Return the reference to the child of this Entry to navigate in the tree.

Returns:

A reference to the child of this Entry

getData

protected ClusKernel getData()

Getter for the data. It is the real object, that means side effects are possible!

Returns:

A reference to the data Kernel in this entry.

See Also:

Kernel

getParentEntry

public Entry getParentEntry()

setParentEntry

public void setParentEntry(Entry parent)

setChild

public void setChild(Node child)

Setter for the child in this entry. Use to build the tree.

Parameters:

child - The Node that should be a child of this Entry

See Also:

Node

getTimestamp

public long getTimestamp()

Return the current timestamp.

Returns:

The current timestamp.

emptyBuffer

protected ClusKernel emptyBuffer(long currentTime,
                     double negLambda)

Clear the buffer in this entry and return a copy. No side effects are possible (given that the copy constructor of Kernel makes a deep copy).

Returns:

A copy of the buffer.

isEmpty

protected boolean isEmpty()

Check if this Entry is empty or not. An Entry is empty if the data Kernel is empty, since then the buffer has to be empty.

Returns:

true if the data cluster has no data points, false otherwise.

overwriteOldEntry

protected void overwriteOldEntry(Entry newEntry)

Overwrites the LS, SS and weightedN in the data cluster of this Entry to the values of the data cluster in the given Entry, but adds N and classCount of the cluster in the given Entry to the data cluster in this one. This function is useful when the weight of an entry becomes to small, and we want to forget the information of the old points.

Parameters:

newEntry - The cluster that should overwrite the information.

recalculateData

protected void recalculateData()

This functions reads every entry in the child node and calculates the corresponding data Kernel. Timestamps are not changed.

See Also:

data, Kerne

isIrrelevant

protected boolean isIrrelevant(double threshold)

Returns true if this entry is irrelevant with respecto the given threshold. This is done by comparing the weighted N of this Entry to the threshold, if it is smaller, than the entry is deemed to be irrelevant.

Parameters:

threshold - The threshold under which entries at leafs can be erased.

Returns:

True if this entry is deemed irrelevant, false otherwise.

makeOlder

protected void makeOlder(long currentTime,
             double negLambda)

Ages this entrie's data AND buffer according to the given time and aging constant.

Parameters:

currentTime - the current time

negLambda - the aging constant