org.openimaj.ml.clustering.assignment.hard

Class ConstrainedFloatAssigner<DATATYPE>

java.lang.Object

org.openimaj.ml.clustering.assignment.hard.ConstrainedFloatAssigner<DATATYPE>

Type Parameters:

DATATYPE - the primitive array datatype which represents a centroid of this cluster.

All Implemented Interfaces:

Assigner<DATATYPE>, HardAssigner<DATATYPE,float[],IntFloatPair>

@Reference(author={"Y. Cai","W. Tong","L. Yang","A. G. Hauptmann"},
           title="Constrained Keypoint Quantization: Towards Better Bag-of-Words Model for Large-scale Multimedia Retrieval",
           type=Inproceedings,
           year="2012",
           booktitle="ACM International Conference on Multimedia Retrieval",
           customData={"location","Hong Kong, China"})
public class ConstrainedFloatAssigner<DATATYPE>
extends Object
implements HardAssigner<DATATYPE,float[],IntFloatPair>

An assigner that wraps another hard assigner and only produces valid assignments if the closest cluster is within (or outside) of a given threshold distance.

Invalid assignments are marked by a cluster id of -1, and (if applicable) distance of Float.NaN. Users of this class must check the assignments and filter as necessary.

Author:

Jonathon Hare (jsh2@ecs.soton.ac.uk)

Constructor Summary

Constructors

Constructor and Description
ConstrainedFloatAssigner(HardAssigner<DATATYPE,float[],IntFloatPair> internalAssigner, float threshold) Construct the ConstrainedFloatAssigner with the given assigner and threshold.
ConstrainedFloatAssigner(HardAssigner<DATATYPE,float[],IntFloatPair> internalAssigner, float threshold, boolean greater) Construct the ConstrainedFloatAssigner with the given assigner and threshold.

Method Summary

Methods

Modifier and Type	Method and Description
int	assign(DATATYPE data) Assign a single point to a cluster.
int[]	assign(DATATYPE[] data) Assign data to a cluster.
IntFloatPair	assignDistance(DATATYPE data) Assign a single point to a cluster.
void	assignDistance(DATATYPE[] data, int[] indices, float[] distances) Assign data to clusters.
int	numDimensions() Get the number of dimensions of the vectors.
int	size() The number of centroids or unique ids that can be generated.

Methods inherited from class java.lang.Object

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

Constructor Detail

ConstrainedFloatAssigner

public ConstrainedFloatAssigner(HardAssigner<DATATYPE,float[],IntFloatPair> internalAssigner,
                        float threshold)

Construct the ConstrainedFloatAssigner with the given assigner and threshold. Assignments will be rejected if the distance given by the internal assigner are greater than the threshold.

Parameters:

internalAssigner - the internal assigner for computing distances.

threshold - the threshold at which assignments are rejected.

ConstrainedFloatAssigner

public ConstrainedFloatAssigner(HardAssigner<DATATYPE,float[],IntFloatPair> internalAssigner,
                        float threshold,
                        boolean greater)

Construct the ConstrainedFloatAssigner with the given assigner and threshold. The greater flag determines if assignments should be rejected if the distance generated by the internal assigner is greater than the threshold (false) or less than the threshold (true).

Parameters:

internalAssigner - the internal assigner for computing distances.

threshold - the threshold at which assignments are rejected.

greater - if true distances less than the threshold are rejected; if false then distances greater than the threshold are rejected.

Method Detail

assign

public int[] assign(DATATYPE[] data)

Description copied from interface: HardAssigner

Assign data to a cluster.

Specified by:

assign in interface HardAssigner<DATATYPE,float[],IntFloatPair>

Parameters:

data - the data.

Returns:

The cluster indices which the data was assigned to.

assign

public int assign(DATATYPE data)

Description copied from interface: HardAssigner

Assign a single point to a cluster.

Specified by:

assign in interface HardAssigner<DATATYPE,float[],IntFloatPair>

Parameters:

data - datum to assign.

Returns:

the cluster index.

assignDistance

public void assignDistance(DATATYPE[] data,
                  int[] indices,
                  float[] distances)

Description copied from interface: HardAssigner

Assign data to clusters. The results are returned in the indices and distances arrays. The return arrays must have the same length as the data array.

Specified by:

assignDistance in interface HardAssigner<DATATYPE,float[],IntFloatPair>

Parameters:

data - the data.

indices - the cluster index for each data point.

distances - the distance to the closest cluster for each data point.

assignDistance

public IntFloatPair assignDistance(DATATYPE data)

Description copied from interface: HardAssigner

Assign a single point to a cluster.

Specified by:

assignDistance in interface HardAssigner<DATATYPE,float[],IntFloatPair>

Parameters:

data - point to assign.

Returns:

the cluster index and distance.

size

public int size()

Description copied from interface: HardAssigner

The number of centroids or unique ids that can be generated.

Specified by:

size in interface HardAssigner<DATATYPE,float[],IntFloatPair>

Returns:

The number of centroids or unique ids that can be generated.

numDimensions

public int numDimensions()

Description copied from interface: Assigner

Get the number of dimensions of the vectors.

Specified by:

numDimensions in interface Assigner<DATATYPE>

Returns:

the number of dimensions