moa.classifiers.lazy.neighboursearch.kdtrees

Class MedianOfWidestDimension

java.lang.Object

moa.classifiers.lazy.neighboursearch.kdtrees.KDTreeNodeSplitter

moa.classifiers.lazy.neighboursearch.kdtrees.MedianOfWidestDimension

All Implemented Interfaces:

Serializable

public class MedianOfWidestDimension
extends KDTreeNodeSplitter

The class that splits a KDTree node based on the median value of a dimension in which the node's points have the widest spread.

For more information see also:

Jerome H. Friedman, Jon Luis Bentley, Raphael Ari Finkel (1977). An Algorithm for Finding Best Matches in Logarithmic Expected Time. ACM Transactions on Mathematics Software. 3(3):209-226.

BibTeX:

 @article{Friedman1977,
    author = {Jerome H. Friedman and Jon Luis Bentley and Raphael Ari Finkel},
    journal = {ACM Transactions on Mathematics Software},
    month = {September},
    number = {3},
    pages = {209-226},
    title = {An Algorithm for Finding Best Matches in Logarithmic Expected Time},
    volume = {3},
    year = {1977}
 }
 

Version:

$Revision: 8034 $

Author:

Ashraf M. Kibriya (amk14[at-the-rate]cs[dot]waikato[dot]ac[dot]nz)

See Also:

Serialized Form

Field Summary

Fields inherited from class moa.classifiers.lazy.neighboursearch.kdtrees.KDTreeNodeSplitter

m_EuclideanDistance, m_Instances, m_InstList, m_NormalizeNodeWidth, MAX, MIN, WIDTH

Constructor Summary

Constructors

Constructor and Description
MedianOfWidestDimension()

Method Summary

Methods

Modifier and Type	Method and Description
String	globalInfo() Returns a string describing this nearest neighbour search algorithm.
protected int	partition(int attIdx, int[] index, int l, int r) Partitions the instances around a pivot.
int	select(int attIdx, int[] indices, int left, int right, int k) Implements computation of the kth-smallest element according to Manber's "Introduction to Algorithms".
void	splitNode(KDTreeNode node, int numNodesCreated, double[][] nodeRanges, double[][] universe) Splits a node into two based on the median value of the dimension in which the points have the widest spread.

Methods inherited from class moa.classifiers.lazy.neighboursearch.kdtrees.KDTreeNodeSplitter

correctlyInitialized, getOptions, listOptions, setEuclideanDistanceFunction, setInstanceList, setInstances, setNodeWidthNormalization, setOptions, widestDim

Methods inherited from class java.lang.Object

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

Constructor Detail

MedianOfWidestDimension

public MedianOfWidestDimension()

Method Detail

globalInfo

public String globalInfo()

Returns a string describing this nearest neighbour search algorithm.

Returns:

a description of the algorithm for displaying in the explorer/experimenter gui

splitNode

public void splitNode(KDTreeNode node,
             int numNodesCreated,
             double[][] nodeRanges,
             double[][] universe)
               throws Exception

Splits a node into two based on the median value of the dimension in which the points have the widest spread. After splitting two new nodes are created and correctly initialised. And, node.left and node.right are set appropriately.

Specified by:

splitNode in class KDTreeNodeSplitter

Parameters:

node - The node to split.

numNodesCreated - The number of nodes that so far have been created for the tree, so that the newly created nodes are assigned correct/meaningful node numbers/ids.

nodeRanges - The attributes' range for the points inside the node that is to be split.

universe - The attributes' range for the whole point-space.

Throws:

Exception - If there is some problem in splitting the given node.

partition

protected int partition(int attIdx,
            int[] index,
            int l,
            int r)

Partitions the instances around a pivot. Used by quicksort and kthSmallestValue.

Parameters:

attIdx - The attribution/dimension based on which the instances should be partitioned.

index - The master index array containing indices of the instances.

l - The begining index of the portion of master index array that should be partitioned.

r - The end index of the portion of master index array that should be partitioned.

Returns:

the index of the middle element

select

public int select(int attIdx,
         int[] indices,
         int left,
         int right,
         int k)

Implements computation of the kth-smallest element according to Manber's "Introduction to Algorithms".

Parameters:

attIdx - The dimension/attribute of the instances in which to find the kth-smallest element.

indices - The master index array containing indices of the instances.

left - The begining index of the portion of the master index array in which to find the kth-smallest element.

right - The end index of the portion of the master index array in which to find the kth-smallest element.

k - The value of k

Returns:

The index of the kth-smallest element