weka.core.metrics
Class WeightedEuclidean

java.lang.Object
  extended byweka.core.metrics.Metric
      extended byweka.core.metrics.LearnableMetric
          extended byweka.core.metrics.WeightedEuclidean
All Implemented Interfaces:
java.lang.Cloneable, OptionHandler, java.io.Serializable

public class WeightedEuclidean
extends LearnableMetric
implements OptionHandler

WeightedEuclidean class Implements weighted euclidean distance metric

See Also:
Serialized Form

Field Summary
static int CONVERSION_EXPONENTIAL
           
static int CONVERSION_LAPLACIAN
          We can have different ways of converting from distance to similarity
static int CONVERSION_UNIT
           
protected  int m_conversionType
          The method of converting, by default laplacian
protected  MetricLearner m_metricLearner
          A metric learner responsible for training the parameters of the metric
static Tag[] TAGS_CONVERSION
           
 
Fields inherited from class weka.core.metrics.LearnableMetric
m_attrWeights, m_classifier, m_classifierClassName, m_classifierRequiresNominalClass, m_numPosDiffInstances, m_posNegDiffInstanceRatio, m_trainable
 
Fields inherited from class weka.core.metrics.Metric
m_attrIdxs, m_classIndex, m_numAttributes
 
Constructor Summary
WeightedEuclidean()
          Create a default new metric
WeightedEuclidean(int numAttributes)
          Create a new metric.
WeightedEuclidean(int[] _attrIdxs)
          Creates a new metric which takes specified attributes.
 
Method Summary
 void buildMetric(Instances data)
          Create a new metric for operating on specified instances
 void buildMetric(int numAttributes)
          Generates a new Metric.
 void buildMetric(int numAttributes, java.lang.String[] options)
          Generates a new Metric.
 java.lang.Object clone()
          Create a copy of this metric
 Instance createDiffInstance(Instance instance1, Instance instance2)
          Create an instance with features corresponding to dot-product components of the two given instances
protected  Instance createDiffInstanceNonSparse(Instance instance1, Instance instance2)
          Create a nonsparse instance with features corresponding to dot-product components of the two given instances
protected  SparseInstance createDiffInstanceSparse(SparseInstance instance1, SparseInstance instance2)
          Create a sparse instance with features corresponding to dot-product components of the two given instances
protected  Instance createDiffInstanceSparseNonSparse(SparseInstance instance1, Instance instance2)
          Create an instance with features corresponding to dot-product components of the two given instances
 double distance(Instance instance1, Instance instance2)
          Returns a distance value between two instances.
 double distanceInternal(Instance instance1, Instance instance2)
          Returns a distance value between two instances.
 double distanceNonSparse(Instance instance1, Instance instance2)
          Returns a distance value between non-sparse instances without using the weights
 double distanceNonSparseNonWeighted(Instance instance1, Instance instance2)
          Returns a distance value between non-sparse instances (or a non-sparse instance and a sparse instance) without using the weights
 double distanceNonWeighted(Instance instance1, Instance instance2)
          Returns a distance value between two instances.
 double distanceSparse(SparseInstance instance1, SparseInstance instance2)
          Returns a distance value between two sparse instances.
 double distanceSparseNonSparse(SparseInstance instance1, Instance instance2)
          Returns a distance value between a non-sparse instance and a sparse instance
 double distanceSparseNonSparseNonWeighted(SparseInstance instance1, Instance instance2)
          Returns a distance value between a non-sparse instance and a sparse instance
 double distanceSparseNonWeighted(SparseInstance instance1, SparseInstance instance2)
          Returns a distance value between two sparse instances without using the weights.
 Instance getCentroidInstance(Instances instances, boolean fastMode, boolean normalized)
          Given a cluster of instances, return the centroid of that cluster
 SelectedTag getConversionType()
          return the type of distance to similarity conversion
 double[] getGradients(Instance instance1, Instance instance2)
          Get the values of the partial derivates for the metric components for a particular instance pair
 MetricLearner getMetricLearner()
          Get the distance metric learner
protected  java.lang.String getMetricLearnerSpec()
          Gets the classifier specification string, which contains the class name of the classifier and any options to the classifier
 java.lang.String[] getOptions()
          Gets the current settings of WeightedEuclideanP.
 boolean isDistanceBased()
          The computation of a metric can be either based on distance, or on similarity
 void learnMetric(Instances data)
          Train the metric
 java.util.Enumeration listOptions()
          Returns an enumeration describing the available options.
static void main(java.lang.String[] args)
           
 void resetMetric()
          Reset all values that have been learned
 void setConversionType(SelectedTag conversionType)
          Set the type of distance to similarity conversion.
 void setMetricLearner(MetricLearner metricLearner)
          Set the distance metric learner
 void setOptions(java.lang.String[] options)
          Parses a given list of options.
 double similarity(Instance instance1, Instance instance2)
          Returns a similarity estimate between two instances.
 double similarityNonWeighted(Instance instance1, Instance instance2)
          Returns a similarity estimate between two instances without using the weights.
 
Methods inherited from class weka.core.metrics.LearnableMetric
getExternal, getNumPosDiffInstances, getPosNegDiffInstanceRatio, getTrainable, getWeights, meanOrMode, normalizeInstanceWeighted, setExternal, setNumPosDiffInstances, setPosNegDiffInstanceRatio, setTrainable, setWeights, useClassifier, useNoClassifier, usesClassifier
 
Methods inherited from class weka.core.metrics.Metric
forName, getAttrIdxs, getAttrIdxsWithoutLastClass, getAttrIndxs, getClassIndex, getNumAttributes, length, normalizeInstance, setAttrIdxs, setAttrIdxs, setClassIndex
 
Methods inherited from class java.lang.Object
equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
 

Field Detail

CONVERSION_LAPLACIAN

public static final int CONVERSION_LAPLACIAN
We can have different ways of converting from distance to similarity

See Also:
Constant Field Values

CONVERSION_UNIT

public static final int CONVERSION_UNIT
See Also:
Constant Field Values

CONVERSION_EXPONENTIAL

public static final int CONVERSION_EXPONENTIAL
See Also:
Constant Field Values

TAGS_CONVERSION

public static final Tag[] TAGS_CONVERSION

m_conversionType

protected int m_conversionType
The method of converting, by default laplacian


m_metricLearner

protected MetricLearner m_metricLearner
A metric learner responsible for training the parameters of the metric

Constructor Detail

WeightedEuclidean

public WeightedEuclidean(int numAttributes)
                  throws java.lang.Exception
Create a new metric.

Parameters:
numAttributes - the number of attributes that the metric will work on

WeightedEuclidean

public WeightedEuclidean()
Create a default new metric


WeightedEuclidean

public WeightedEuclidean(int[] _attrIdxs)
                  throws java.lang.Exception
Creates a new metric which takes specified attributes.

Parameters:
_attrIdxs - An array containing attribute indeces that will be used in the metric
Method Detail

resetMetric

public void resetMetric()
                 throws java.lang.Exception
Reset all values that have been learned

Specified by:
resetMetric in class LearnableMetric
Throws:
java.lang.Exception

buildMetric

public void buildMetric(int numAttributes)
                 throws java.lang.Exception
Generates a new Metric. Has to initialize all fields of the metric with default values.

Specified by:
buildMetric in class Metric
Parameters:
numAttributes - the number of attributes that the metric will work on
Throws:
java.lang.Exception - if the distance metric has not been generated successfully.

buildMetric

public void buildMetric(int numAttributes,
                        java.lang.String[] options)
                 throws java.lang.Exception
Generates a new Metric. Has to initialize all fields of the metric with default values

Specified by:
buildMetric in class Metric
Parameters:
options - an array of options suitable for passing to setOptions. May be null.
numAttributes - the number of attributes that the metric will work on
Throws:
java.lang.Exception - if the distance metric has not been generated successfully.

buildMetric

public void buildMetric(Instances data)
                 throws java.lang.Exception
Create a new metric for operating on specified instances

Specified by:
buildMetric in class Metric
Parameters:
data - instances that the metric will be used on
Throws:
java.lang.Exception

distance

public double distance(Instance instance1,
                       Instance instance2)
                throws java.lang.Exception
Returns a distance value between two instances.

Specified by:
distance in class Metric
Parameters:
instance1 - First instance.
instance2 - Second instance.
Throws:
java.lang.Exception - if distance could not be estimated.

distanceInternal

public double distanceInternal(Instance instance1,
                               Instance instance2)
                        throws java.lang.Exception
Returns a distance value between two instances.

Parameters:
instance1 - First instance.
instance2 - Second instance.
Throws:
java.lang.Exception - if distance could not be estimated.

distanceSparse

public double distanceSparse(SparseInstance instance1,
                             SparseInstance instance2)
                      throws java.lang.Exception
Returns a distance value between two sparse instances.

Parameters:
instance1 - First sparse instance.
instance2 - Second sparse instance.
Throws:
java.lang.Exception - if distance could not be estimated.

distanceSparseNonSparse

public double distanceSparseNonSparse(SparseInstance instance1,
                                      Instance instance2)
                               throws java.lang.Exception
Returns a distance value between a non-sparse instance and a sparse instance

Parameters:
instance1 - sparse instance.
instance2 - sparse instance.
Throws:
java.lang.Exception - if distance could not be estimated.

distanceNonSparse

public double distanceNonSparse(Instance instance1,
                                Instance instance2)
                         throws java.lang.Exception
Returns a distance value between non-sparse instances without using the weights

Parameters:
instance1 - non-sparse instance.
instance2 - non-sparse instance.
Throws:
java.lang.Exception - if distance could not be estimated.

distanceNonWeighted

public double distanceNonWeighted(Instance instance1,
                                  Instance instance2)
                           throws java.lang.Exception
Returns a distance value between two instances.

Specified by:
distanceNonWeighted in class Metric
Parameters:
instance1 - First instance.
instance2 - Second instance.
Throws:
java.lang.Exception - if distance could not be estimated.

distanceSparseNonWeighted

public double distanceSparseNonWeighted(SparseInstance instance1,
                                        SparseInstance instance2)
                                 throws java.lang.Exception
Returns a distance value between two sparse instances without using the weights.

Parameters:
instance1 - First sparse instance.
instance2 - Second sparse instance.
Throws:
java.lang.Exception - if distance could not be estimated.

distanceSparseNonSparseNonWeighted

public double distanceSparseNonSparseNonWeighted(SparseInstance instance1,
                                                 Instance instance2)
                                          throws java.lang.Exception
Returns a distance value between a non-sparse instance and a sparse instance

Parameters:
instance1 - sparse instance.
instance2 - sparse instance.
Throws:
java.lang.Exception - if distance could not be estimated.

distanceNonSparseNonWeighted

public double distanceNonSparseNonWeighted(Instance instance1,
                                           Instance instance2)
                                    throws java.lang.Exception
Returns a distance value between non-sparse instances (or a non-sparse instance and a sparse instance) without using the weights

Parameters:
instance1 - non-sparse instance.
instance2 - non-sparse instance.
Throws:
java.lang.Exception - if distance could not be estimated.

similarity

public double similarity(Instance instance1,
                         Instance instance2)
                  throws java.lang.Exception
Returns a similarity estimate between two instances. Similarity is obtained by inverting the distance value using one of three methods: CONVERSION_LAPLACIAN, CONVERSION_EXPONENTIAL, CONVERSION_UNIT.

Specified by:
similarity in class Metric
Parameters:
instance1 - First instance.
instance2 - Second instance.
Throws:
java.lang.Exception - if similarity could not be estimated.

similarityNonWeighted

public double similarityNonWeighted(Instance instance1,
                                    Instance instance2)
                             throws java.lang.Exception
Returns a similarity estimate between two instances without using the weights.

Specified by:
similarityNonWeighted in class Metric
Parameters:
instance1 - First instance.
instance2 - Second instance.
Throws:
java.lang.Exception - if similarity could not be estimated.

getGradients

public double[] getGradients(Instance instance1,
                             Instance instance2)
                      throws java.lang.Exception
Get the values of the partial derivates for the metric components for a particular instance pair

Specified by:
getGradients in class LearnableMetric
Parameters:
instance1 - the first instance
instance2 - the first instance
Throws:
java.lang.Exception

learnMetric

public void learnMetric(Instances data)
                 throws java.lang.Exception
Train the metric

Specified by:
learnMetric in class LearnableMetric
Throws:
java.lang.Exception

setMetricLearner

public void setMetricLearner(MetricLearner metricLearner)
Set the distance metric learner

Parameters:
metricLearner - the metric learner

getMetricLearner

public MetricLearner getMetricLearner()
Get the distance metric learner


createDiffInstance

public Instance createDiffInstance(Instance instance1,
                                   Instance instance2)
Create an instance with features corresponding to dot-product components of the two given instances

Specified by:
createDiffInstance in class LearnableMetric
Parameters:
instance1 - first instance
instance2 - second instance

createDiffInstanceSparse

protected SparseInstance createDiffInstanceSparse(SparseInstance instance1,
                                                  SparseInstance instance2)
Create a sparse instance with features corresponding to dot-product components of the two given instances

Parameters:
instance1 - first sparse instance
instance2 - second sparse instance

createDiffInstanceSparseNonSparse

protected Instance createDiffInstanceSparseNonSparse(SparseInstance instance1,
                                                     Instance instance2)
Create an instance with features corresponding to dot-product components of the two given instances

Parameters:
instance1 - first sparse instance
instance2 - second non-sparse instance

createDiffInstanceNonSparse

protected Instance createDiffInstanceNonSparse(Instance instance1,
                                               Instance instance2)
Create a nonsparse instance with features corresponding to dot-product components of the two given instances

Parameters:
instance1 - first nonsparse instance
instance2 - second nonsparse instance

setConversionType

public void setConversionType(SelectedTag conversionType)
Set the type of distance to similarity conversion. Values other than CONVERSION_LAPLACIAN, CONVERSION_UNIT, or CONVERSION_EXPONENTIAL will be ignored


getConversionType

public SelectedTag getConversionType()
return the type of distance to similarity conversion

Returns:
one of CONVERSION_LAPLACIAN, CONVERSION_UNIT, or CONVERSION_EXPONENTIAL

isDistanceBased

public boolean isDistanceBased()
The computation of a metric can be either based on distance, or on similarity

Specified by:
isDistanceBased in class Metric

getCentroidInstance

public Instance getCentroidInstance(Instances instances,
                                    boolean fastMode,
                                    boolean normalized)
Given a cluster of instances, return the centroid of that cluster

Specified by:
getCentroidInstance in class LearnableMetric
Parameters:
instances - objects belonging to a cluster
fastMode - whether fast mode should be used for SparseInstances
normalized - normalize centroids for SPKMeans
Returns:
a centroid instance for the given cluster

setOptions

public void setOptions(java.lang.String[] options)
                throws java.lang.Exception
Parses a given list of options. Valid options are:

-N
Normalize the euclidean distance by vectors lengths -E
Use exponential conversion from distance to similarity (default laplacian conversion)

-U
Use unit conversion from similarity to distance (dist=1-sim) (default laplacian conversion)

-R
The metric is trainable and will be trained using the current MetricLearner (default non-trainable)

Specified by:
setOptions in interface OptionHandler
Parameters:
options - the list of options as an array of strings
Throws:
java.lang.Exception - if an option is not supported

getMetricLearnerSpec

protected java.lang.String getMetricLearnerSpec()
Gets the classifier specification string, which contains the class name of the classifier and any options to the classifier

Returns:
the classifier string.

listOptions

public java.util.Enumeration listOptions()
Returns an enumeration describing the available options.

Specified by:
listOptions in interface OptionHandler
Returns:
an enumeration of all the available options.

getOptions

public java.lang.String[] getOptions()
Gets the current settings of WeightedEuclideanP.

Specified by:
getOptions in interface OptionHandler
Returns:
an array of strings suitable for passing to setOptions()

clone

public java.lang.Object clone()
Create a copy of this metric

Overrides:
clone in class LearnableMetric

main

public static void main(java.lang.String[] args)