org.apache.mahout.math
Interface Vector

All Superinterfaces:
Cloneable
All Known Implementing Classes:
AbstractMatrix.TransposeViewVector, AbstractVector, Centroid, ConstantVector, DelegatingVector, DenseVector, MatrixSlice, MatrixVectorView, NamedVector, PermutedVectorView, RandomAccessSparseVector, SequentialAccessSparseVector, VectorView, WeightedVector

public interface Vector
extends Cloneable

The basic interface including numerous convenience functions

NOTE: All implementing classes must have a constructor that takes an int for cardinality and a no-arg constructor that can be used for marshalling the Writable instance

NOTE: Implementations may choose to reuse the Vector.Element in the Iterable methods


Nested Class Summary
static interface Vector.Element
          A holder for information about a specific item in the Vector.
 
Method Summary
 double aggregate(DoubleDoubleFunction aggregator, DoubleFunction map)
          Examples speak louder than words: aggregate(plus, pow(2)) is another way to say getLengthSquared(), aggregate(max, abs) is norm(Double.POSITIVE_INFINITY).
 double aggregate(Vector other, DoubleDoubleFunction aggregator, DoubleDoubleFunction combiner)
          Generalized inner product - take two vectors, iterate over them both, using the combiner to combine together (and possibly map in some way) each pair of values, which are then aggregated with the previous accumulated value in the combiner.
 Iterable<Vector.Element> all()
           
 String asFormatString()
           
 Vector assign(double value)
          Assign the value to all elements of the receiver
 Vector assign(double[] values)
          Assign the values to the receiver
 Vector assign(DoubleDoubleFunction f, double y)
          Apply the function to each element of the receiver, using the y value as the second argument of the DoubleDoubleFunction
 Vector assign(DoubleFunction function)
          Apply the function to each element of the receiver
 Vector assign(Vector other)
          Assign the other vector values to the receiver
 Vector assign(Vector other, DoubleDoubleFunction function)
          Apply the function to each element of the receiver and the corresponding element of the other argument
 Vector clone()
          Return a copy of the recipient
 Matrix cross(Vector other)
          Return the cross product of the receiver and the other vector
 Vector divide(double x)
          Return a new vector containing the values of the recipient divided by the argument
 double dot(Vector x)
          Return the dot product of the recipient and the argument
 double get(int index)
          Return the value at the given index
 double getDistanceSquared(Vector v)
          Get the square of the distance between this vector and the other vector.
 Vector.Element getElement(int index)
          Return an object of Vector.Element representing an element of this Vector.
 double getIteratorAdvanceCost()
          Gets an estimate of the cost (in number of operations) it takes to advance an iterator through the nonzero elements of this vector.
 double getLengthSquared()
          Return the sum of squares of all elements in the vector.
 double getLookupCost()
          Gets an estimate of the cost (in number of operations) it takes to lookup a random element in this vector.
 int getNumNondefaultElements()
          Return the number of values in the recipient which are not the default value.
 int getNumNonZeroElements()
          Return the number of non zero elements in the vector.
 double getQuick(int index)
          Return the value at the given index, without checking bounds
 void incrementQuick(int index, double increment)
          Increment the value at the given index by the given value.
 boolean isAddConstantTime()
          Return true iff adding a new (nonzero) element takes constant time for this vector.
 boolean isDense()
           
 boolean isSequentialAccess()
           
 Vector like()
          Return an empty vector of the same underlying class as the receiver
 Vector logNormalize()
          Return a new vector containing the log(1 + entry)/ L_2 norm values of the recipient
 Vector logNormalize(double power)
          Return a new Vector with a normalized value calculated as log_power(1 + entry)/ L_power norm.
 double maxValue()
           
 int maxValueIndex()
           
 void mergeUpdates(OrderedIntDoubleMapping updates)
          Merge a set of (index, value) pairs into the vector.
 Vector minus(Vector x)
          Return a new vector containing the element by element difference of the recipient and the argument
 double minValue()
           
 int minValueIndex()
           
 Iterable<Vector.Element> nonZeroes()
           
 double norm(double power)
          Return the k-norm of the vector.
 Vector normalize()
          Return a new vector containing the normalized (L_2 norm) values of the recipient
 Vector normalize(double power)
          Return a new Vector containing the normalized (L_power norm) values of the recipient.
 Vector plus(double x)
          Return a new vector containing the sum of each value of the recipient and the argument
 Vector plus(Vector x)
          Return a new vector containing the element by element sum of the recipient and the argument
 void set(int index, double value)
          Set the value at the given index
 void setQuick(int index, double value)
          Set the value at the given index, without checking bounds
 int size()
          Return the cardinality of the recipient (the maximum number of values)
 Vector times(double x)
          Return a new vector containing the product of each value of the recipient and the argument
 Vector times(Vector x)
          Return a new vector containing the element-wise product of the recipient and the argument
 Vector viewPart(int offset, int length)
          Return a new vector containing the subset of the recipient
 double zSum()
          Return the sum of all the elements of the receiver
 

Method Detail

asFormatString

String asFormatString()
Returns:
a formatted String suitable for output

assign

Vector assign(double value)
Assign the value to all elements of the receiver

Parameters:
value - a double value
Returns:
the modified receiver

assign

Vector assign(double[] values)
Assign the values to the receiver

Parameters:
values - a double[] of values
Returns:
the modified receiver
Throws:
CardinalityException - if the cardinalities differ

assign

Vector assign(Vector other)
Assign the other vector values to the receiver

Parameters:
other - a Vector
Returns:
the modified receiver
Throws:
CardinalityException - if the cardinalities differ

assign

Vector assign(DoubleFunction function)
Apply the function to each element of the receiver

Parameters:
function - a DoubleFunction to apply
Returns:
the modified receiver

assign

Vector assign(Vector other,
              DoubleDoubleFunction function)
Apply the function to each element of the receiver and the corresponding element of the other argument

Parameters:
other - a Vector containing the second arguments to the function
function - a DoubleDoubleFunction to apply
Returns:
the modified receiver
Throws:
CardinalityException - if the cardinalities differ

assign

Vector assign(DoubleDoubleFunction f,
              double y)
Apply the function to each element of the receiver, using the y value as the second argument of the DoubleDoubleFunction

Parameters:
f - a DoubleDoubleFunction to be applied
y - a double value to be argument to the function
Returns:
the modified receiver

size

int size()
Return the cardinality of the recipient (the maximum number of values)

Returns:
an int

isDense

boolean isDense()
Returns:
true iff this implementation should be considered dense -- that it explicitly represents every value

isSequentialAccess

boolean isSequentialAccess()
Returns:
true iff this implementation should be considered to be iterable in index order in an efficient way. In particular this implies that all() and nonZeroes() ()} return elements in ascending order by index.

clone

Vector clone()
Return a copy of the recipient

Returns:
a new Vector

all

Iterable<Vector.Element> all()

nonZeroes

Iterable<Vector.Element> nonZeroes()

getElement

Vector.Element getElement(int index)
Return an object of Vector.Element representing an element of this Vector. Useful when designing new iterator types.

Parameters:
index - Index of the Vector.Element required
Returns:
The Vector.Element Object

mergeUpdates

void mergeUpdates(OrderedIntDoubleMapping updates)
Merge a set of (index, value) pairs into the vector.

Parameters:
updates - an ordered mapping of indices to values to be merged in.

divide

Vector divide(double x)
Return a new vector containing the values of the recipient divided by the argument

Parameters:
x - a double value
Returns:
a new Vector

dot

double dot(Vector x)
Return the dot product of the recipient and the argument

Parameters:
x - a Vector
Returns:
a new Vector
Throws:
CardinalityException - if the cardinalities differ

get

double get(int index)
Return the value at the given index

Parameters:
index - an int index
Returns:
the double at the index
Throws:
IndexException - if the index is out of bounds

getQuick

double getQuick(int index)
Return the value at the given index, without checking bounds

Parameters:
index - an int index
Returns:
the double at the index

like

Vector like()
Return an empty vector of the same underlying class as the receiver

Returns:
a Vector

minus

Vector minus(Vector x)
Return a new vector containing the element by element difference of the recipient and the argument

Parameters:
x - a Vector
Returns:
a new Vector
Throws:
CardinalityException - if the cardinalities differ

normalize

Vector normalize()
Return a new vector containing the normalized (L_2 norm) values of the recipient

Returns:
a new Vector

normalize

Vector normalize(double power)
Return a new Vector containing the normalized (L_power norm) values of the recipient.

See http://en.wikipedia.org/wiki/Lp_space

Technically, when 0 < power < 1, we don't have a norm, just a metric, but we'll overload this here.

Also supports power == 0 (number of non-zero elements) and power = Double.POSITIVE_INFINITY (max element). Again, see the Wikipedia page for more info

Parameters:
power - The power to use. Must be >= 0. May also be Double.POSITIVE_INFINITY. See the Wikipedia link for more on this.
Returns:
a new Vector x such that norm(x, power) == 1

logNormalize

Vector logNormalize()
Return a new vector containing the log(1 + entry)/ L_2 norm values of the recipient

Returns:
a new Vector

logNormalize

Vector logNormalize(double power)
Return a new Vector with a normalized value calculated as log_power(1 + entry)/ L_power norm.

Parameters:
power - The power to use. Must be > 1. Cannot be Double.POSITIVE_INFINITY.
Returns:
a new Vector

norm

double norm(double power)
Return the k-norm of the vector.

See http://en.wikipedia.org/wiki/Lp_space

Technically, when 0 > power < 1, we don't have a norm, just a metric, but we'll overload this here. Also supports power == 0 (number of non-zero elements) and power = Double.POSITIVE_INFINITY (max element). Again, see the Wikipedia page for more info.

Parameters:
power - The power to use.
See Also:
normalize(double)

minValue

double minValue()
Returns:
The minimum value in the Vector

minValueIndex

int minValueIndex()
Returns:
The index of the minimum value

maxValue

double maxValue()
Returns:
The maximum value in the Vector

maxValueIndex

int maxValueIndex()
Returns:
The index of the maximum value

plus

Vector plus(double x)
Return a new vector containing the sum of each value of the recipient and the argument

Parameters:
x - a double
Returns:
a new Vector

plus

Vector plus(Vector x)
Return a new vector containing the element by element sum of the recipient and the argument

Parameters:
x - a Vector
Returns:
a new Vector
Throws:
CardinalityException - if the cardinalities differ

set

void set(int index,
         double value)
Set the value at the given index

Parameters:
index - an int index into the receiver
value - a double value to set
Throws:
IndexException - if the index is out of bounds

setQuick

void setQuick(int index,
              double value)
Set the value at the given index, without checking bounds

Parameters:
index - an int index into the receiver
value - a double value to set

incrementQuick

void incrementQuick(int index,
                    double increment)
Increment the value at the given index by the given value.

Parameters:
index - an int index into the receiver
increment - sets the value at the given index to value + increment;

getNumNondefaultElements

int getNumNondefaultElements()
Return the number of values in the recipient which are not the default value. For instance, for a sparse vector, this would be the number of non-zero values.

Returns:
an int

getNumNonZeroElements

int getNumNonZeroElements()
Return the number of non zero elements in the vector.

Returns:
an int

times

Vector times(double x)
Return a new vector containing the product of each value of the recipient and the argument

Parameters:
x - a double argument
Returns:
a new Vector

times

Vector times(Vector x)
Return a new vector containing the element-wise product of the recipient and the argument

Parameters:
x - a Vector argument
Returns:
a new Vector
Throws:
CardinalityException - if the cardinalities differ

viewPart

Vector viewPart(int offset,
                int length)
Return a new vector containing the subset of the recipient

Parameters:
offset - an int offset into the receiver
length - the cardinality of the desired result
Returns:
a new Vector
Throws:
CardinalityException - if the length is greater than the cardinality of the receiver
IndexException - if the offset is negative or the offset+length is outside of the receiver

zSum

double zSum()
Return the sum of all the elements of the receiver

Returns:
a double

cross

Matrix cross(Vector other)
Return the cross product of the receiver and the other vector

Parameters:
other - another Vector
Returns:
a Matrix

aggregate

double aggregate(DoubleDoubleFunction aggregator,
                 DoubleFunction map)
Examples speak louder than words: aggregate(plus, pow(2)) is another way to say getLengthSquared(), aggregate(max, abs) is norm(Double.POSITIVE_INFINITY). To sum all of the positive values, aggregate(plus, max(0)).

Parameters:
aggregator - used to combine the current value of the aggregation with the result of map.apply(nextValue)
map - a function to apply to each element of the vector in turn before passing to the aggregator
Returns:
the final aggregation

aggregate

double aggregate(Vector other,
                 DoubleDoubleFunction aggregator,
                 DoubleDoubleFunction combiner)

Generalized inner product - take two vectors, iterate over them both, using the combiner to combine together (and possibly map in some way) each pair of values, which are then aggregated with the previous accumulated value in the combiner.

Example: dot(other) could be expressed as aggregate(other, Plus, Times), and kernelized inner products (which are symmetric on the indices) work similarly.

Parameters:
other - a vector to aggregate in combination with
aggregator - function we're aggregating with; fa
combiner - function we're combining with; fc
Returns:
the final aggregation; if r0 = fc(this[0], other[0]), ri = fa(r_{i-1}, fc(this[i], other[i])) for all i > 0

getLengthSquared

double getLengthSquared()
Return the sum of squares of all elements in the vector. Square root of this value is the length of the vector.


getDistanceSquared

double getDistanceSquared(Vector v)
Get the square of the distance between this vector and the other vector.


getLookupCost

double getLookupCost()
Gets an estimate of the cost (in number of operations) it takes to lookup a random element in this vector.


getIteratorAdvanceCost

double getIteratorAdvanceCost()
Gets an estimate of the cost (in number of operations) it takes to advance an iterator through the nonzero elements of this vector.


isAddConstantTime

boolean isAddConstantTime()
Return true iff adding a new (nonzero) element takes constant time for this vector.



Copyright © 2008–2014 The Apache Software Foundation. All rights reserved.