org.apache.ignite.ml.knn

Class NNClassificationModel

java.lang.Object

org.apache.ignite.ml.knn.NNClassificationModel

All Implemented Interfaces:

Serializable, Function<Vector,Double>, Exportable<KNNModelFormat>, IgniteFunction<Vector,Double>, Model<Vector,Double>

Direct Known Subclasses:

ANNClassificationModel, KNNClassificationModel

public abstract class NNClassificationModel
extends Object
implements Model<Vector,Double>, Exportable<KNNModelFormat>

Common methods and fields for all kNN and aNN models to predict label based on neighbours' labels.

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
protected DistanceMeasure	distanceMeasure Distance measure.
protected int	k Amount of nearest neighbors.
protected NNStrategy	stgy kNN strategy.

Constructor Summary

Constructors

Constructor and Description
NNClassificationModel()

Method Summary

Modifier and Type	Method and Description
protected LabeledVectorSet<Double,LabeledVector>	buildLabeledDatasetOnListOfVectors(List<LabeledVector> neighborsFromPartitions)
protected void	copyParametersFrom(NNClassificationModel mdl) Sets parameters from other model to this model.
boolean	equals(Object obj)
protected double	getClassVoteForVector(NNStrategy stgy, double distance)
protected double	getClassWithMaxVotes(Map<Double,Double> clsVotes)
DistanceMeasure	getDistanceMeasure()
protected @NotNull TreeMap<Double,Set<Integer>>	getDistances(Vector v, LabeledVectorSet<Double,LabeledVector> trainingData) Computes distances between given vector and each vector in training dataset.
protected @NotNull LabeledVector[]	getKClosestVectors(LabeledVectorSet<Double,LabeledVector> trainingData, TreeMap<Double,Set<Integer>> distanceIdxPairs) Iterates along entries in distance map and fill the resulting k-element array.
int	hashCode()
protected void	putDistanceIdxPair(Map<Double,Set<Integer>> distanceIdxPairs, int i, double distance)
abstract <P> void	saveModel(Exporter<KNNModelFormat,P> exporter, P path) Save model by the given path.
String	toString()
String	toString(boolean pretty)
NNClassificationModel	withDistanceMeasure(DistanceMeasure distanceMeasure) Set up parameter of the NN model.
NNClassificationModel	withK(int k) Set up parameter of the NN model.
NNClassificationModel	withStrategy(NNStrategy stgy) Set up parameter of the NN model.

Methods inherited from class java.lang.Object

clone, finalize, getClass, notify, notifyAll, wait, wait, wait

Methods inherited from interface org.apache.ignite.ml.Model

combine

Methods inherited from interface java.util.function.Function

andThen, apply, compose, identity

Field Detail

k

protected int k

Amount of nearest neighbors.

distanceMeasure

protected DistanceMeasure distanceMeasure

Distance measure.

stgy

protected NNStrategy stgy

kNN strategy.

Constructor Detail

NNClassificationModel

public NNClassificationModel()

Method Detail

withK

public NNClassificationModel withK(int k)

Set up parameter of the NN model.

Parameters:

k - Amount of nearest neighbors.

Returns:

Model.

withStrategy

public NNClassificationModel withStrategy(NNStrategy stgy)

Set up parameter of the NN model.

Parameters:

stgy - Strategy of calculations.

Returns:

Model.

withDistanceMeasure

public NNClassificationModel withDistanceMeasure(DistanceMeasure distanceMeasure)

Set up parameter of the NN model.

Parameters:

distanceMeasure - Distance measure.

Returns:

Model.

buildLabeledDatasetOnListOfVectors

protected LabeledVectorSet<Double,LabeledVector> buildLabeledDatasetOnListOfVectors(List<LabeledVector> neighborsFromPartitions)

getKClosestVectors

@NotNull
protected @NotNull LabeledVector[] getKClosestVectors(LabeledVectorSet<Double,LabeledVector> trainingData,
                                                               TreeMap<Double,Set<Integer>> distanceIdxPairs)

Iterates along entries in distance map and fill the resulting k-element array.

Parameters:

trainingData - The training data.

distanceIdxPairs - The distance map.

Returns:

K-nearest neighbors.

getDistances

@NotNull
protected @NotNull TreeMap<Double,Set<Integer>> getDistances(Vector v,
                                                                      LabeledVectorSet<Double,LabeledVector> trainingData)

Computes distances between given vector and each vector in training dataset.

Parameters:

v - The given vector.

trainingData - The training dataset.

Returns:

Key - distanceMeasure from given features before features with idx stored in value. Value is presented with Set because there can be a few vectors with the same distance.

putDistanceIdxPair

protected void putDistanceIdxPair(Map<Double,Set<Integer>> distanceIdxPairs,
                                  int i,
                                  double distance)

getClassWithMaxVotes

protected double getClassWithMaxVotes(Map<Double,Double> clsVotes)

getClassVoteForVector

protected double getClassVoteForVector(NNStrategy stgy,
                                       double distance)

getDistanceMeasure

public DistanceMeasure getDistanceMeasure()

hashCode

public int hashCode()

Overrides:

hashCode in class Object

equals

public boolean equals(Object obj)

Overrides:

equals in class Object

toString

public String toString()

Overrides:

toString in class Object

toString

public String toString(boolean pretty)

Specified by:

toString in interface Model<Vector,Double>

Parameters:

pretty - Use pretty mode.

copyParametersFrom

protected void copyParametersFrom(NNClassificationModel mdl)

Sets parameters from other model to this model.

Parameters:

mdl - Model.

saveModel

public abstract <P> void saveModel(Exporter<KNNModelFormat,P> exporter,
                                   P path)

Description copied from interface: Exportable

Save model by the given path.

Specified by:

saveModel in interface Exportable<KNNModelFormat>

Parameters:

exporter - Exporter.

path - Path to saved model.