Package weka.core.neighboursearch
Class KDTree
java.lang.Object
weka.core.neighboursearch.NearestNeighbourSearch
weka.core.neighboursearch.KDTree
- All Implemented Interfaces:
Serializable,AdditionalMeasureProducer,OptionHandler,RevisionHandler,TechnicalInformationHandler
Class implementing the KDTree search algorithm for nearest neighbour search.
The connection to dataset is only a reference. For the tree structure the indexes are stored in an array.
Building the tree:
If a node has <maximal-inst-number> (option -L) instances no further splitting is done. Also if the split would leave one side empty, the branch is not split any further even if the instances in the resulting node are more than <maximal-inst-number> instances.
**PLEASE NOTE:** The algorithm can not handle missing values, so it is advisable to run ReplaceMissingValues filter if there are any missing values in the dataset.
For more information see:
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.
Andrew Moore (1991). A tutorial on kd-trees. BibTeX:
The connection to dataset is only a reference. For the tree structure the indexes are stored in an array.
Building the tree:
If a node has <maximal-inst-number> (option -L) instances no further splitting is done. Also if the split would leave one side empty, the branch is not split any further even if the instances in the resulting node are more than <maximal-inst-number> instances.
**PLEASE NOTE:** The algorithm can not handle missing values, so it is advisable to run ReplaceMissingValues filter if there are any missing values in the dataset.
For more information see:
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.
Andrew Moore (1991). A tutorial on kd-trees. 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}
}
@techreport{Moore1991,
author = {Andrew Moore},
booktitle = {University of Cambridge Computer Laboratory Technical Report No. 209},
howpublished = {Extract from PhD Thesis},
title = {A tutorial on kd-trees},
year = {1991},
HTTP = {Available from http://www.autonlab.org/autonweb/14665.html}
}
Valid options are:
-S <classname and options> Node splitting method to use. (default: weka.core.neighboursearch.kdtrees.SlidingMidPointOfWidestSide)
-W <value> Set minimal width of a box (default: 1.0E-2).
-L Maximal number of instances in a leaf (default: 40).
-N Normalizing will be done (Select dimension for split, with normalising to universe).
- Version:
- $Revision: 14831 $
- Author:
- Gabi Schmidberger (gabi[at-the-rate]cs[dot]waikato[dot]ac[dot]nz), Malcolm Ware (mfw4[at-the-rate]cs[dot]waikato[dot]ac[dot]nz), Ashraf M. Kibriya (amk14[at-the-rate]cs[dot]waikato[dot]ac[dot]nz)
- See Also:
-
Field Summary
Fields -
Constructor Summary
Constructors -
Method Summary
Modifier and TypeMethodDescriptionvoidaddInstanceInfo(Instance instance) Adds one instance to KDTree loosly.voidassignSubToCenters(KDTreeNode node, Instances centers, int[] centList, int[] assignments) Assigns instances of this node to center.voidcenterInstances(Instances centers, int[] assignments, double pc) Assigns instances to centers using KDTree.Returns an enumeration of the additional measure names.returns the distance function currently in use.double[]Returns the distances to the kNearest or 1 nearest neighbour currently found with either the kNearestNeighbours or the nearestNeighbour method.intGet the maximum number of instances in a leaf.doublegetMeasure(String additionalMeasureName) Returns the value of the named measure.doubleGets the minimum relative box width.Returns the splitting method currently in use to split the nodes of the KDTree.booleanGets the normalize flag.String[]Gets the current settings of KDtree.Returns the revision string.Returns an instance of a TechnicalInformation object, containing detailed information about the technical background of this class, e.g., paper reference or book this class is based on.Returns a string describing this nearest neighbour search algorithm.kNearestNeighbours(Instance target, int k) Returns the k nearest neighbours of the supplied instance.Returns an enumeration describing the available options.Tip text for this property.doubleReturns the depth of the tree.doubleReturns the number of leaves.doubleReturns the size of the tree.Tip text for this property.nearestNeighbour(Instance target) Returns the nearest neighbour of the supplied target instance.Returns the tip text for this property.Tip text for this property.voidsets the distance function to use for nearest neighbour search.voidsetInstances(Instances instances) Builds the KDTree on the given set of instances.voidsetMaxInstInLeaf(int i) Sets the maximum number of instances in a leaf.voidsetMeasurePerformance(boolean measurePerformance) Sets whether to calculate the performance statistics or not.voidsetMinBoxRelWidth(double i) Sets the minimum relative box width.voidsetNodeSplitter(KDTreeNodeSplitter splitter) Sets the splitting method to use to split the nodes of the KDTree.voidsetNormalizeNodeWidth(boolean n) Sets the flag for normalizing the widths of a KDTree Node by the width of the dimension in the universe.voidsetOptions(String[] options) Parses a given list of options.voidAdds one instance to the KDTree.Methods inherited from class weka.core.neighboursearch.NearestNeighbourSearch
combSort11, distanceFunctionTipText, getInstances, getMeasurePerformance, getPerformanceStats, measurePerformanceTipText, quickSort
-
Field Details
-
MIN
public static final int MINThe index of MIN value in attributes' range array.- See Also:
-
MAX
public static final int MAXThe index of MAX value in attributes' range array.- See Also:
-
WIDTH
public static final int WIDTHThe index of WIDTH (MAX-MIN) value in attributes' range array.- See Also:
-
-
Constructor Details
-
KDTree
public KDTree()Creates a new instance of KDTree. -
KDTree
Creates a new instance of KDTree. It also builds the tree on supplied set of Instances.- Parameters:
insts- The instances/points on which the BallTree should be built on.
-
-
Method Details
-
getTechnicalInformation
Returns an instance of a TechnicalInformation object, containing detailed information about the technical background of this class, e.g., paper reference or book this class is based on.- Specified by:
getTechnicalInformationin interfaceTechnicalInformationHandler- Returns:
- the technical information about this class
-
kNearestNeighbours
Returns the k nearest neighbours of the supplied instance. >k neighbours are returned if there are more than one neighbours at the kth boundary.- Specified by:
kNearestNeighboursin classNearestNeighbourSearch- Parameters:
target- The instance to find the nearest neighbours for.k- The number of neighbours to find.- Returns:
- The k nearest neighbours (or >k if more there are than one neighbours at the kth boundary).
- Throws:
Exception- if the nearest neighbour could not be found.
-
nearestNeighbour
Returns the nearest neighbour of the supplied target instance.- Specified by:
nearestNeighbourin classNearestNeighbourSearch- Parameters:
target- The instance to find the nearest neighbour for.- Returns:
- The nearest neighbour from among the previously supplied training instances.
- Throws:
Exception- if the neighbours could not be found.
-
getDistances
Returns the distances to the kNearest or 1 nearest neighbour currently found with either the kNearestNeighbours or the nearestNeighbour method.- Specified by:
getDistancesin classNearestNeighbourSearch- Returns:
- array containing the distances of the nearestNeighbours. The length and ordering of the array is the same as that of the instances returned by nearestNeighbour functions.
- Throws:
Exception- if called before calling kNearestNeighbours or nearestNeighbours.
-
setInstances
Builds the KDTree on the given set of instances.- Overrides:
setInstancesin classNearestNeighbourSearch- Parameters:
instances- The insts on which the KDTree is to be built.- Throws:
Exception- If some error occurs while building the KDTree
-
update
Adds one instance to the KDTree. This updates the KDTree structure to take into account the newly added training instance.- Specified by:
updatein classNearestNeighbourSearch- Parameters:
instance- the instance to be added. Usually the newly added instance in the training set.- Throws:
Exception- If the instance cannot be added.
-
addInstanceInfo
Adds one instance to KDTree loosly. It only changes the ranges in EuclideanDistance, and does not affect the structure of the KDTree.- Overrides:
addInstanceInfoin classNearestNeighbourSearch- Parameters:
instance- the new instance. Usually this is the test instance supplied to update the range of attributes in the distance function.
-
measureTreeSize
public double measureTreeSize()Returns the size of the tree.- Returns:
- the size of the tree
-
measureNumLeaves
public double measureNumLeaves()Returns the number of leaves.- Returns:
- the number of leaves
-
measureMaxDepth
public double measureMaxDepth()Returns the depth of the tree.- Returns:
- The depth of the tree
-
enumerateMeasures
Returns an enumeration of the additional measure names.- Specified by:
enumerateMeasuresin interfaceAdditionalMeasureProducer- Overrides:
enumerateMeasuresin classNearestNeighbourSearch- Returns:
- an enumeration of the measure names
-
getMeasure
Returns the value of the named measure.- Specified by:
getMeasurein interfaceAdditionalMeasureProducer- Overrides:
getMeasurein classNearestNeighbourSearch- Parameters:
additionalMeasureName- the name of the measure to query for its value.- Returns:
- The value of the named measure
- Throws:
IllegalArgumentException- If the named measure is not supported.
-
setMeasurePerformance
public void setMeasurePerformance(boolean measurePerformance) Sets whether to calculate the performance statistics or not.- Overrides:
setMeasurePerformancein classNearestNeighbourSearch- Parameters:
measurePerformance- Should be true if performance statistics are to be measured.
-
centerInstances
Assigns instances to centers using KDTree.- Parameters:
centers- the current centersassignments- the centerindex for each instancepc- the threshold value for pruning.- Throws:
Exception- If there is some problem assigning instances to centers.
-
assignSubToCenters
public void assignSubToCenters(KDTreeNode node, Instances centers, int[] centList, int[] assignments) throws Exception Assigns instances of this node to center. Center to be assign to is decided by the distance function.- Parameters:
node- The KDTreeNode whose instances are to be assigned.centers- all the input centerscentList- the list of centers to work withassignments- index list of last assignments- Throws:
Exception- If there is error assigning the instances.
-
minBoxRelWidthTipText
Tip text for this property.- Returns:
- the tip text for this property
-
setMinBoxRelWidth
public void setMinBoxRelWidth(double i) Sets the minimum relative box width.- Parameters:
i- the minimum relative box width
-
getMinBoxRelWidth
public double getMinBoxRelWidth()Gets the minimum relative box width.- Returns:
- the minimum relative box width
-
maxInstInLeafTipText
Tip text for this property.- Returns:
- the tip text for this property
-
setMaxInstInLeaf
public void setMaxInstInLeaf(int i) Sets the maximum number of instances in a leaf.- Parameters:
i- the maximum number of instances in a leaf
-
getMaxInstInLeaf
public int getMaxInstInLeaf()Get the maximum number of instances in a leaf.- Returns:
- the maximum number of instances in a leaf
-
normalizeNodeWidthTipText
Tip text for this property.- Returns:
- the tip text for this property
-
setNormalizeNodeWidth
public void setNormalizeNodeWidth(boolean n) Sets the flag for normalizing the widths of a KDTree Node by the width of the dimension in the universe.- Parameters:
n- true to use normalizing.
-
getNormalizeNodeWidth
public boolean getNormalizeNodeWidth()Gets the normalize flag.- Returns:
- True if normalizing
-
getDistanceFunction
returns the distance function currently in use.- Overrides:
getDistanceFunctionin classNearestNeighbourSearch- Returns:
- the distance function
-
setDistanceFunction
sets the distance function to use for nearest neighbour search.- Overrides:
setDistanceFunctionin classNearestNeighbourSearch- Parameters:
df- the distance function to use- Throws:
Exception- if not EuclideanDistance
-
nodeSplitterTipText
Returns the tip text for this property.- Returns:
- tip text for this property suitable for displaying in the explorer/experimenter gui
-
getNodeSplitter
Returns the splitting method currently in use to split the nodes of the KDTree.- Returns:
- The KDTreeNodeSplitter currently in use.
-
setNodeSplitter
Sets the splitting method to use to split the nodes of the KDTree.- Parameters:
splitter- The KDTreeNodeSplitter to use.
-
globalInfo
Returns a string describing this nearest neighbour search algorithm.- Overrides:
globalInfoin classNearestNeighbourSearch- Returns:
- a description of the algorithm for displaying in the explorer/experimenter gui
-
listOptions
Returns an enumeration describing the available options.- Specified by:
listOptionsin interfaceOptionHandler- Overrides:
listOptionsin classNearestNeighbourSearch- Returns:
- an enumeration of all the available options.
-
setOptions
Parses a given list of options. Valid options are:-S <classname and options> Node splitting method to use. (default: weka.core.neighboursearch.kdtrees.SlidingMidPointOfWidestSide)
-W <value> Set minimal width of a box (default: 1.0E-2).
-L Maximal number of instances in a leaf (default: 40).
-N Normalizing will be done (Select dimension for split, with normalising to universe).
- Specified by:
setOptionsin interfaceOptionHandler- Overrides:
setOptionsin classNearestNeighbourSearch- Parameters:
options- the list of options as an array of strings- Throws:
Exception- if an option is not supported
-
getOptions
Gets the current settings of KDtree.- Specified by:
getOptionsin interfaceOptionHandler- Overrides:
getOptionsin classNearestNeighbourSearch- Returns:
- an array of strings suitable for passing to setOptions
-
getRevision
Returns the revision string.- Specified by:
getRevisionin interfaceRevisionHandler- Returns:
- the revision
-