javatools.datatypes
Class SparseVector

java.lang.Object
  javatools.datatypes.SparseVector

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable

public class SparseVector
extends java.lang.Object
implements java.io.Serializable, java.lang.Cloneable

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). This class represents a Sparse Vector, i.e. a vector that has only few non-zero entries. Supports special (space-)optimized routines for binary vectors. Supports k-means with eucledian, cosine or other distances.

See Also:

Serialized Form

Nested Class Summary

static interface

SparseVector.Distance Computes a distance of two vectors

Field Summary

static SparseVector.Distance	cosineDistance Computes the cosine distance
static SparseVector.Distance	eucledianDistance Computes the eucledian distance

Constructor Summary

SparseVector(double[] v, int[] d)
Creates a Sparse Vector from (sorted) dimensions and values

SparseVector(double label, double... values)
Creates a SparseVector from actual values

SparseVector(double l, int[] d)
Creates a binary Sparse Vector from an unsorted array of dimensions and a label Zero-entries and duplicates in the dimension array will be ignored.

SparseVector(double l, int[] d, double[] v)
Creates a Sparse Vector from (sorted) dimensions, (modifiable) values and a label

SparseVector(double l, java.util.List<java.lang.Integer> dimensions)
Creates a binary Sparse Vector from an (unsorted) list of dimensions and a label

SparseVector(SparseVector v)
Duplicates a vector

SparseVector(java.lang.String s)
Creates a Sparse Vector from a SVM-light-like input string of the form label dim:val ...

Method Summary

SparseVector	add(SparseVector v) Adds a vector to this one
char	charLabel() Returns a char label for the vector
SparseVector	clone() Clones a vector
SparseVector	compress() Kicks out 0-entries (without affecting the vector in a mathematical sense)
double	cosine(SparseVector v) Returns the cosine of this vector with another vector
double	eucledianDistance(SparseVector v) Returns the distance to another vector
double	get(int i) Retrieves the i-th value
int	index(int dimension) Returns the index for a dimension (or -1)
boolean	isBinary() Tells whether this vector is binary
static void	kMeans(SparseVector[] dots, SparseVector[] centers) Does a simple K-means
static void	kMeans(SparseVector[] dots, SparseVector[] centers, SparseVector.Distance distanceFunction, double epsilon, int iterations) Does a simple K-means until the number of iterations is exceeded
double	l2norm() Returns this vector's two-norm
static void	main(java.lang.String[] argv) Test method
SparseVector	multiply(double r) Multiplies this vector by a scalar
java.util.Iterator<java.lang.Integer>	nonZeroIndices() Gives an iterator over the non-zero indices
int	size() Returns the number of non-zero elements.
double	sprod(SparseVector v) Returns the scalar product (dot product) of this vector with another one.
double	squaredl2norm() Returns this vector's squared two-norm
java.lang.String	toString() Returns a String representation in an SVM-light-like format label dim:val ...
static java.lang.String	visualize(java.util.List<SparseVector> vectors) Visualizes two-dimensional vectors
static java.lang.String	visualize(java.util.List<SparseVector> vectors1, java.util.List<SparseVector> vectors2) Visualizes two-dimensional vectors
static java.lang.String	visualize(SparseVector... vectors) Visualizes two-dimensional vectors
static java.lang.String	visualize(SparseVector[] vectors1, SparseVector[] vectors2) Visualizes two-dimensional vectors

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, wait, wait, wait

Field Detail

eucledianDistance

public static final SparseVector.Distance eucledianDistance

Computes the eucledian distance

cosineDistance

public static final SparseVector.Distance cosineDistance

Computes the cosine distance

Constructor Detail

SparseVector

public SparseVector(SparseVector v)

Duplicates a vector

SparseVector

public SparseVector(double l,
                    java.util.List<java.lang.Integer> dimensions)

Creates a binary Sparse Vector from an (unsorted) list of dimensions and a label

SparseVector

public SparseVector(double l,
                    int[] d)

Creates a binary Sparse Vector from an unsorted array of dimensions and a label Zero-entries and duplicates in the dimension array will be ignored. Left here as a legacy from LEILA

SparseVector

public SparseVector(double[] v,
                    int[] d)

Creates a Sparse Vector from (sorted) dimensions and values

SparseVector

public SparseVector(double l,
                    int[] d,
                    double[] v)

Creates a Sparse Vector from (sorted) dimensions, (modifiable) values and a label

SparseVector

public SparseVector(double label,
                    double... values)

Creates a SparseVector from actual values

SparseVector

public SparseVector(java.lang.String s)

Creates a Sparse Vector from a SVM-light-like input string of the form label dim:val ... # comments

Method Detail

compress

public SparseVector compress()

Kicks out 0-entries (without affecting the vector in a mathematical sense)

toString

public java.lang.String toString()

Returns a String representation in an SVM-light-like format label dim:val ...

Overrides:

toString in class java.lang.Object

sprod

public double sprod(SparseVector v)

Returns the scalar product (dot product) of this vector with another one. This code is optimized by speed, not by size

isBinary

public boolean isBinary()

Tells whether this vector is binary

size

public int size()

Returns the number of non-zero elements.

squaredl2norm

public double squaredl2norm()

Returns this vector's squared two-norm

l2norm

public double l2norm()

Returns this vector's two-norm

cosine

public double cosine(SparseVector v)

Returns the cosine of this vector with another vector

get

public double get(int i)

Retrieves the i-th value

visualize

public static java.lang.String visualize(SparseVector... vectors)

Visualizes two-dimensional vectors

visualize

public static java.lang.String visualize(java.util.List<SparseVector> vectors1,
                                         java.util.List<SparseVector> vectors2)

Visualizes two-dimensional vectors

visualize

public static java.lang.String visualize(SparseVector[] vectors1,
                                         SparseVector[] vectors2)

Visualizes two-dimensional vectors

visualize

public static java.lang.String visualize(java.util.List<SparseVector> vectors)

Visualizes two-dimensional vectors

charLabel

public char charLabel()

Returns a char label for the vector

kMeans

public static void kMeans(SparseVector[] dots,
                          SparseVector[] centers)

Does a simple K-means

kMeans

public static void kMeans(SparseVector[] dots,
                          SparseVector[] centers,
                          SparseVector.Distance distanceFunction,
                          double epsilon,
                          int iterations)

Does a simple K-means until the number of iterations is exceeded

eucledianDistance

public double eucledianDistance(SparseVector v)

Returns the distance to another vector

index

public int index(int dimension)

Returns the index for a dimension (or -1)

add

public SparseVector add(SparseVector v)

Adds a vector to this one

multiply

public SparseVector multiply(double r)

Multiplies this vector by a scalar

nonZeroIndices

public java.util.Iterator<java.lang.Integer> nonZeroIndices()

Gives an iterator over the non-zero indices

clone

public SparseVector clone()

Clones a vector

Overrides:

clone in class java.lang.Object

main

public static void main(java.lang.String[] argv)
                 throws java.lang.Exception

Test method

Throws:

java.lang.Exception