de.lmu.ifi.dbs.elki.data

Class VectorUtil

java.lang.Object

de.lmu.ifi.dbs.elki.data.VectorUtil

public final class VectorUtil
extends Object

Utility functions for use with vectors. Note: obviously, many functions are class methods or database related.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	VectorUtil.SortDBIDsBySingleDimension Compare number vectors by a single dimension.
static class	VectorUtil.SortVectorsBySingleDimension Compare number vectors by a single dimension.

Constructor Summary

Constructors

Modifier	Constructor and Description
private	VectorUtil() Fake constructor.

Method Summary

Methods

Modifier and Type	Method and Description
static double	angle(NumberVector<?> v1, NumberVector<?> v2, NumberVector<?> o) Compute the angle between two vectors.
static double	angle(NumberVector<?> v1, NumberVector<?> v2, Vector o) Compute the angle between two vectors.
static double	angleSparse(SparseNumberVector<?> v1, SparseNumberVector<?> v2) Compute the angle for sparse vectors.
static Vector	computeMedoid(Relation<? extends NumberVector<?>> relation, DBIDs sample) Compute medoid for a given subset.
static double	cosAngle(NumberVector<?> v1, NumberVector<?> v2) Compute the absolute cosine of the angle between two vectors.
static double[]	fastTimes(Matrix mat, NumberVector<?> v) This is an ugly hack, but we don't want to have the Matrix class depend on NumberVector.
static DoubleMinMax	getRangeDouble(NumberVector<?> vec) Return the range across all dimensions.
static double	minCosAngle(SpatialComparable v1, SpatialComparable v2) Compute the minimum angle between two rectangles.
static <V extends NumberVector<?>> V	project(V v, BitSet selectedAttributes, NumberVector.Factory<V,?> factory) Provides a new NumberVector as a projection on the specified attributes.
static <V extends NumberVector<?>> V	randomVector(NumberVector.Factory<V,?> factory, int dim) Produce a new vector based on random numbers in [0:1].
static <V extends NumberVector<?>> V	randomVector(NumberVector.Factory<V,?> factory, int dim, Random r) Produce a new vector based on random numbers in [0:1].
static double	scalarProduct(NumberVector<?> d1, NumberVector<?> d2) Provides the scalar product (inner product) of this and the given DoubleVector.

Methods inherited from class java.lang.Object

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

Constructor Detail

VectorUtil

private VectorUtil()

Fake constructor. Do not instantiate, use static methods.

Method Detail

getRangeDouble

public static DoubleMinMax getRangeDouble(NumberVector<?> vec)

Return the range across all dimensions. Useful in particular for time series.

Parameters:

vec - Vector to process.

Returns:

[min, max]

randomVector

public static <V extends NumberVector<?>> V randomVector(NumberVector.Factory<V,?> factory,
                                         int dim,
                                         Random r)

Produce a new vector based on random numbers in [0:1].

Type Parameters:

V - vector type

Parameters:

factory - Vector factory

dim - desired dimensionality

r - Random generator

Returns:

new instance

randomVector

public static <V extends NumberVector<?>> V randomVector(NumberVector.Factory<V,?> factory,
                                         int dim)

Produce a new vector based on random numbers in [0:1].

Type Parameters:

V - vector type

Parameters:

factory - Vector factory

dim - desired dimensionality

Returns:

new instance

angleSparse

public static double angleSparse(SparseNumberVector<?> v1,
                 SparseNumberVector<?> v2)

Compute the angle for sparse vectors.

Parameters:

v1 - First vector

v2 - Second vector

Returns:

angle

angle

public static double angle(NumberVector<?> v1,
           NumberVector<?> v2,
           Vector o)

Compute the angle between two vectors.

Parameters:

v1 - first vector

v2 - second vector

o - Origin

Returns:

Angle

angle

public static double angle(NumberVector<?> v1,
           NumberVector<?> v2,
           NumberVector<?> o)

Compute the angle between two vectors.

Parameters:

v1 - first vector

v2 - second vector

o - Origin

Returns:

Angle

cosAngle

public static double cosAngle(NumberVector<?> v1,
              NumberVector<?> v2)

Compute the absolute cosine of the angle between two vectors. To convert it to radians, use Math.acos(angle)!

Parameters:

v1 - first vector

v2 - second vector

Returns:

Angle

minCosAngle

public static double minCosAngle(SpatialComparable v1,
                 SpatialComparable v2)

Compute the minimum angle between two rectangles.

Parameters:

v1 - first rectangle

v2 - second rectangle

Returns:

Angle

scalarProduct

public static double scalarProduct(NumberVector<?> d1,
                   NumberVector<?> d2)

Provides the scalar product (inner product) of this and the given DoubleVector.

Parameters:

d1 - the first vector to compute the scalar product for

d2 - the second vector to compute the scalar product for

Returns:

the scalar product (inner product) of this and the given DoubleVector

computeMedoid

public static Vector computeMedoid(Relation<? extends NumberVector<?>> relation,
                   DBIDs sample)

Compute medoid for a given subset.

Parameters:

relation - Relation to process

sample - Sample set

Returns:

Medoid vector

fastTimes

public static double[] fastTimes(Matrix mat,
                 NumberVector<?> v)

This is an ugly hack, but we don't want to have the Matrix class depend on NumberVector. Maybe a future version will no longer need this.

Parameters:

mat - Matrix

v - Vector

Returns:

mat * v, as double array.

project

public static <V extends NumberVector<?>> V project(V v,
                                    BitSet selectedAttributes,
                                    NumberVector.Factory<V,?> factory)

Provides a new NumberVector as a projection on the specified attributes.

Type Parameters:

V - Vector type

Parameters:

v - a NumberVector to project

selectedAttributes - the attributes selected for projection

factory - Vector factory

Returns:

a new NumberVector as a projection on the specified attributes