FrequencyVector

java.lang.Object
- javatools.datatypes.FrequencyVector<T,V>

```
public class FrequencyVector<T,V extends java.lang.Number & java.lang.Comparable<V>>
extends java.lang.Object
```
This class is part of the Java Tools (see http://mpii.de/yago-naga/javatools). It is licensed under the Creative Commons Attribution License (see http://creativecommons.org/licenses/by/3.0) by the YAGO-NAGA team (see http://mpii.de/yago-naga). Fabian M. Suchanek, Milan Vojnovic, Dinan Gunawardena "Social Tags: Meaning and Suggestions" (pdf, bib) ACM Conference on Information and Knowledge Management (CIKM 2008) The class represents a vector of terms with their frequencies. See the paper for explanations. Also provides Wilson interval computation.

Author:

Fabian M. Suchanek

Field Summary

Fields
Modifier and Type	Field and Description
`protected java.util.Map<T,V>`	`data` Holds the vector itself
`protected double`	`max` Holds the maximum value of all elements
`protected double`	`norm` Holds the L2 norm of this vector
`protected java.util.List<T>`	`sortedTerms` Holds the terms of the vector sorted by decreasing frequency
`protected double`	`sum` Holds the sum of all elements

Constructor Summary

Constructors
Constructor and Description

FrequencyVector(java.util.Map<T,V> applications)
Constructs a frequency vector.

Constructors
Constructor and Description
`FrequencyVector(java.util.Map<T,V> applications)` Constructs a frequency vector.

Method Summary

All Methods Static Methods Instance Methods Concrete Methods
Modifier and Type	Method and Description
`double`	`averagePrecision(java.util.Collection<T> groundTruth)` Computes the average precision (MAP)
`double`	`cosine(FrequencyVector<T,?> other)` Computes the cosine to another vector
`double`	`cosine(FrequencyVector<T,?> other, java.util.Collection<T> intersection)` Computes the cosine to another vector, if the intersection is already available
`double`	`doubleValueFor(T term)` Returns the frequency for a term as double (or 0)
`boolean`	`equals(java.lang.Object obj)`
`int`	`firstTiePos()` Returns the first position in sortedTerms that has equal number of applications to its successor
`double`	`fuzzyPrecisionWithRespectTo(FrequencyVector<T,V> trueFrequencies)` Computes the fuzzy Precision
`double`	`fuzzyRecallWithRespectTo(FrequencyVector<T,V> trueFrequencies)` Computes the fuzzy recall
`int`	`hashCode()`
`java.util.Set<T>`	`intersection(FrequencyVector<T,?> other)` Computes the common supports
`static void`	`main(java.lang.String[] args)` Test
`double`	`max()` Returns the maximum of elements
`FrequencyVector<T,java.lang.Double>`	`maxNormalized()` Max-Normalizes this vector
`double`	`maxNormalizedValueFor(T term)` Returns the frequency for a term, divided by the maximum (or 0)
`double`	`ndcg2WithRespectToGain(FrequencyVector<T,?> trueFrequencies)` Computes the NDCG with respect to a gain, with weighting 2^x
`double`	`ndcgWithRespectToGain(FrequencyVector<T,?> trueFrequencies)` Computes the NDCG with respect to a gain
`double`	`norm()` Returns the L2 norm of elements
`FrequencyVector<T,java.lang.Double>`	`normalized()` Normalizes this vector
`FrequencyVector<T,java.lang.Double>`	`normalizedMeanWith(FrequencyVector<T,V> other)` Computes the mean vector of this vector and the other one
`double`	`normalizedValueFor(T term)` Returns the frequency for a term, divided by the sum (or 0)
`int`	`numTerms()` Returns the number of terms in the support
`double`	`optimalAveragePrecision(FrequencyVector<T,?> trueFrequencies)` Computes the average precision, shuffling subsets to get an optimal value
`double`	`precisionAtKWithRespectTo(java.util.Collection<T> groundTruth, int k)` Computes the standard precision at k
`double`	`precisionAtKWithRespectTo(FrequencyVector<T,?> groundTruth, int k)` Computes the standard precision at k
`double`	`precisionWithRespectTo(java.util.Collection<T> groundTruth)` Computes standard precision
`double`	`precisionWithRespectTo(FrequencyVector<T,?> trueFrequencies)` Computes the standard precision
`double`	`precisionWithRespectToIntersection(java.util.Collection<T> intersection)` Computes the standard precision if the intersection is known
`double`	`recallAtKWithRespectTo(java.util.Collection<T> groundTruth, int k)` Computes the standard recall at k
`double`	`recallAtKWithRespectTo(FrequencyVector<T,V> trueFrequencies, int k)` Computes the standard recall at k
`double`	`recallWithRespectTo(java.util.Collection<T> groundTruth)` Computes the standard recall
`double`	`recallWithRespectTo(java.util.Collection<T> trueSet, java.util.Collection<T> intersection)` Computes standard recall, if the intersection is already available
`double`	`recallWithRespectTo(FrequencyVector<T,?> trueFrequencies)` Computes standard recall
`double`	`recallWithRespectTo(FrequencyVector<T,?> trueFrequencies, java.util.Collection<T> intersection)` Computes standard recall, if the intersection is already available
`double`	`smoothedValueFor(T term)` Returns the frequency for a term, divided by the sum and smoothed
`java.util.List<T>`	`sortedTerms()` Returns the terms sorted by decreasing frequency
`protected static <C,V extends java.lang.Comparable<V>> java.util.List<C>`	`sortedTerms(java.util.Map<C,V> applications)` Returns the terms of the application data sorted by their number of applications
`double`	`sum()` Returns the sum of elements
`T`	`termAtRank(int i)` Returns the term at rank i
`java.util.Collection<T>`	`terms()` Returns the terms.
`java.util.Set<T>`	`topKIntersection(FrequencyVector<T,V> trueFrequencies, int k)` Computes the intersection of the top k elements
`java.lang.String`	`toString()`
`V`	`valueFor(T term)` Returns the frequency for a term (or null)
`double`	`weightedPrecisionAtKWithRespectTo(java.util.Collection<T> groundTruth, int k)` Computes the precision at k to a set, weighted with this vector's frequencies
`double`	`weightedPrecisionAtKWithRespectTo(FrequencyVector<T,?> groundTruth, int k)` Computes the weighted precision at k
`double`	`weightedPrecisionWithRespectTo(java.util.Collection<T> groundTruth)` Computes the standard precision to a set, weighted with this vector's frequencies
`double`	`weightedPrecisionWithRespectTo(FrequencyVector<T,?> trueFrequencies)` Computes the standard precision to a set, weighted with this vector's frequencies
`double`	`weightedRecallAtKWithRespectTo(FrequencyVector<T,V> trueFrequencies, int k)` Computes the standard recall at k, weighted with the true frequencies
`double`	`weightedRecallWithRespectTo(FrequencyVector<T,V> trueFrequencies)` Computes the standard recall, weighted with the true frequencies
`static double[]`	`wilson(int total, int correct)` Computes the Wilson Interval (see http://en.wikipedia.org/wiki/Binomial_proportion_confidence_interval#Wilson_score_interval) Given the total number of events and the number of "correct" events, returns in a double-array in the first component the center of the Wilson interval and in the second component the width of the interval.

Methods inherited from class java.lang.Object
clone, finalize, getClass, notify, notifyAll, wait, wait, wait

Field Detail
- norm
```
protected double norm
```
  Holds the L2 norm of this vector
- sum
```
protected double sum
```
  Holds the sum of all elements
- max
```
protected double max
```
  Holds the maximum value of all elements
- data
```
protected java.util.Map<T,V extends java.lang.Number & java.lang.Comparable<V>> data
```
  Holds the vector itself
- sortedTerms
```
protected java.util.List<T> sortedTerms
```
  Holds the terms of the vector sorted by decreasing frequency

Constructor Detail
- FrequencyVector
```
public FrequencyVector(java.util.Map<T,V> applications)
```
  Constructs a frequency vector. The frequency vector is backed by the map. ZERO-entries are removed!

Method Detail

firstTiePos
```
public int firstTiePos()
```
Returns the first position in sortedTerms that has equal number of applications to its successor

sum
```
public double sum()
```
Returns the sum of elements

norm
```
public double norm()
```
Returns the L2 norm of elements

max
```
public double max()
```
Returns the maximum of elements

termAtRank
```
public T termAtRank(int i)
```
Returns the term at rank i

numTerms
```
public int numTerms()
```
Returns the number of terms in the support

valueFor
```
public V valueFor(T term)
```
Returns the frequency for a term (or null)

doubleValueFor
```
public double doubleValueFor(T term)
```
Returns the frequency for a term as double (or 0)

normalizedValueFor
```
public double normalizedValueFor(T term)
```
Returns the frequency for a term, divided by the sum (or 0)

maxNormalizedValueFor
```
public double maxNormalizedValueFor(T term)
```
Returns the frequency for a term, divided by the maximum (or 0)

smoothedValueFor
```
public double smoothedValueFor(T term)
```
Returns the frequency for a term, divided by the sum and smoothed

sortedTerms
```
public java.util.List<T> sortedTerms()
```
Returns the terms sorted by decreasing frequency

terms
```
public java.util.Collection<T> terms()
```
Returns the terms. This collection may be more efficient for membership checks than the sortedTerms

sortedTerms

protected static <C,V extends java.lang.Comparable<V>> java.util.List<C> sortedTerms(java.util.Map<C,V> applications)

Returns the terms of the application data sorted by their number of applications

toString
```
public java.lang.String toString()
```
Overrides:

toString in class java.lang.Object

hashCode
```
public int hashCode()
```
Overrides:

hashCode in class java.lang.Object

equals

public boolean equals(java.lang.Object obj)

Overrides:: equals in class java.lang.Object

normalized

public FrequencyVector<T,java.lang.Double> normalized()

Normalizes this vector

maxNormalized

public FrequencyVector<T,java.lang.Double> maxNormalized()

Max-Normalizes this vector

intersection

public java.util.Set<T> intersection(FrequencyVector<T,?> other)

Computes the common supports

topKIntersection

public java.util.Set<T> topKIntersection(FrequencyVector<T,V> trueFrequencies,
                                         int k)

Computes the intersection of the top k elements

cosine

public double cosine(FrequencyVector<T,?> other,
                     java.util.Collection<T> intersection)

Computes the cosine to another vector, if the intersection is already available

cosine

public double cosine(FrequencyVector<T,?> other)

Computes the cosine to another vector

precisionAtKWithRespectTo

public double precisionAtKWithRespectTo(java.util.Collection<T> groundTruth,
                                        int k)

Computes the standard precision at k

weightedPrecisionAtKWithRespectTo

public double weightedPrecisionAtKWithRespectTo(java.util.Collection<T> groundTruth,
                                                int k)

Computes the precision at k to a set, weighted with this vector's frequencies

precisionWithRespectTo

public double precisionWithRespectTo(java.util.Collection<T> groundTruth)

Computes standard precision

precisionWithRespectTo

public double precisionWithRespectTo(FrequencyVector<T,?> trueFrequencies)

Computes the standard precision

precisionWithRespectToIntersection
```
public double precisionWithRespectToIntersection(java.util.Collection<T> intersection)
```
Computes the standard precision if the intersection is known

weightedPrecisionWithRespectTo
```
public double weightedPrecisionWithRespectTo(java.util.Collection<T> groundTruth)
```
Computes the standard precision to a set, weighted with this vector's frequencies

weightedPrecisionWithRespectTo
```
public double weightedPrecisionWithRespectTo(FrequencyVector<T,?> trueFrequencies)
```
Computes the standard precision to a set, weighted with this vector's frequencies

precisionAtKWithRespectTo

public double precisionAtKWithRespectTo(FrequencyVector<T,?> groundTruth,
                                        int k)

Computes the standard precision at k

weightedPrecisionAtKWithRespectTo

public double weightedPrecisionAtKWithRespectTo(FrequencyVector<T,?> groundTruth,
                                                int k)

Computes the weighted precision at k

averagePrecision

public double averagePrecision(java.util.Collection<T> groundTruth)

Computes the average precision (MAP)

optimalAveragePrecision
```
public double optimalAveragePrecision(FrequencyVector<T,?> trueFrequencies)
```
Computes the average precision, shuffling subsets to get an optimal value

recallAtKWithRespectTo

public double recallAtKWithRespectTo(java.util.Collection<T> groundTruth,
                                     int k)

Computes the standard recall at k

recallWithRespectTo

public double recallWithRespectTo(java.util.Collection<T> groundTruth)

Computes the standard recall

weightedRecallWithRespectTo
```
public double weightedRecallWithRespectTo(FrequencyVector<T,V> trueFrequencies)
```
Computes the standard recall, weighted with the true frequencies

weightedRecallAtKWithRespectTo

public double weightedRecallAtKWithRespectTo(FrequencyVector<T,V> trueFrequencies,
                                             int k)

Computes the standard recall at k, weighted with the true frequencies

recallWithRespectTo

public double recallWithRespectTo(FrequencyVector<T,?> trueFrequencies)

Computes standard recall

recallWithRespectTo

public double recallWithRespectTo(java.util.Collection<T> trueSet,
                                  java.util.Collection<T> intersection)

Computes standard recall, if the intersection is already available

recallWithRespectTo

public double recallWithRespectTo(FrequencyVector<T,?> trueFrequencies,
                                  java.util.Collection<T> intersection)

Computes standard recall, if the intersection is already available

recallAtKWithRespectTo

public double recallAtKWithRespectTo(FrequencyVector<T,V> trueFrequencies,
                                     int k)

Computes the standard recall at k

ndcgWithRespectToGain

public double ndcgWithRespectToGain(FrequencyVector<T,?> trueFrequencies)

Computes the NDCG with respect to a gain

ndcg2WithRespectToGain
```
public double ndcg2WithRespectToGain(FrequencyVector<T,?> trueFrequencies)
```
Computes the NDCG with respect to a gain, with weighting 2^x

normalizedMeanWith

public FrequencyVector<T,java.lang.Double> normalizedMeanWith(FrequencyVector<T,V> other)

Computes the mean vector of this vector and the other one

fuzzyRecallWithRespectTo

public double fuzzyRecallWithRespectTo(FrequencyVector<T,V> trueFrequencies)

Computes the fuzzy recall

fuzzyPrecisionWithRespectTo

public double fuzzyPrecisionWithRespectTo(FrequencyVector<T,V> trueFrequencies)

Computes the fuzzy Precision

wilson
```
public static double[] wilson(int total,
                              int correct)
```
Computes the Wilson Interval (see http://en.wikipedia.org/wiki/Binomial_proportion_confidence_interval#Wilson_score_interval) Given the total number of events and the number of "correct" events, returns in a double-array in the first component the center of the Wilson interval and in the second component the width of the interval. alpha=95%.

main

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

Test

Class FrequencyVector<T,V extends java.lang.Number & java.lang.Comparable<V>>

Field Summary

Constructor Summary

Method Summary

Methods inherited from class java.lang.Object

Field Detail

norm

sum

max

data

sortedTerms

Constructor Detail

FrequencyVector

Method Detail

firstTiePos

sum

norm

max

termAtRank

numTerms

valueFor

doubleValueFor

normalizedValueFor

maxNormalizedValueFor

smoothedValueFor

sortedTerms

terms

sortedTerms

toString

hashCode

equals

normalized

maxNormalized

intersection

topKIntersection

cosine

cosine

precisionAtKWithRespectTo

weightedPrecisionAtKWithRespectTo

precisionWithRespectTo

precisionWithRespectTo

precisionWithRespectToIntersection

weightedPrecisionWithRespectTo

weightedPrecisionWithRespectTo

precisionAtKWithRespectTo

weightedPrecisionAtKWithRespectTo

averagePrecision

optimalAveragePrecision

recallAtKWithRespectTo

recallWithRespectTo

weightedRecallWithRespectTo

weightedRecallAtKWithRespectTo

recallWithRespectTo

recallWithRespectTo

recallWithRespectTo

recallAtKWithRespectTo

ndcgWithRespectToGain

ndcg2WithRespectToGain

normalizedMeanWith

fuzzyRecallWithRespectTo

fuzzyPrecisionWithRespectTo

wilson

main