Package com.yahoo.tensor

Class IndexedTensor

java.lang.Object

com.yahoo.tensor.IndexedTensor

All Implemented Interfaces:

Tensor

public abstract class IndexedTensor
extends Object
implements Tensor

An indexed (dense) tensor.

Some methods on indexed tensors make use of a standard value order: Cells are ordered by increasing index where dimensions to the right are incremented before indexes to the left, where the order of dimensions are alphabetical by name. In consequence, tensor value ordering is independent of the order in which dimensions are specified, and the values of the right-most dimension are adjacent.

Author:

bratseth

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	IndexedTensor.BoundBuilder	A bound builder can create the double array directly
static class	IndexedTensor.Builder
static interface	IndexedTensor.DirectIndexBuilder
static class	IndexedTensor.Indexes	An array of indexes into this tensor which are able to find the next index in the value order.
final class	IndexedTensor.SubspaceIterator	An iterator over a subspace of this tensor.

Nested classes/interfaces inherited from interface com.yahoo.tensor.Tensor

Tensor.Cell

Method Summary

Modifier and Type	Method	Description
Iterator<Tensor.Cell>	cellIterator()	Returns an iterator over the cells of this in the standard value order.
IndexedTensor.SubspaceIterator	cellIterator(PartialAddress partialAddress, DimensionSizes iterationSizes)	Returns an iterator over all the cells in this tensor which matches the given partial address
Map<TensorAddress,Double>	cells()	Returns an immutable map of the cells of this in no particular order.
DimensionSizes	dimensionSizes()
boolean	equals(Object other)	Returns whether this tensor and the given tensor is mathematically equal: That they have the same dimension *names* and the same content.
abstract double	get(long valueIndex)	Returns the value at the given standard value order index as a double.
double	get(long... indexes)	Returns the value at the given indexes as a double
double	get(TensorAddress address)	Returns the value at this address, or 0.0 if there is no value at this address
abstract float	getFloat(long valueIndex)	Returns the value at the given standard value order index as a float.
float	getFloat(long... indexes)	Returns the value at the given indexes as a float
boolean	has(TensorAddress address)	Returns true if this cell exists
Tensor	remove(Set<TensorAddress> addresses)	Returns a new tensor where existing cells in this tensor have been removed according to the given set of addresses.
long[]	shape()
Iterator<IndexedTensor.SubspaceIterator>	subspaceIterator(Set<String> dimensions)	Returns a subspace iterator having the sizes of the dimensions of this tensor
Iterator<IndexedTensor.SubspaceIterator>	subspaceIterator(Set<String> dimensions, DimensionSizes sizes)	Returns an iterator over value iterators where the outer iterator is over each unique value of the dimensions given and the inner iterator is over each unique value of the rest of the dimensions, in the standard value order
String	toAbbreviatedString(boolean withType, boolean shortForms)	Returns an abbreviated string representation of this tensor suitable for human-readable messages
String	toString()	Returns this tensor on the tensor literal form with type included.
String	toString(boolean withType, boolean shortForms)	Returns this tensor on the tensor literal form.
TensorType	type()
Iterator<Double>	valueIterator()	Returns an iterator over the values of this returned in the standard value order
abstract IndexedTensor	withType(TensorType type)	Returns this tensor with the given type if types are compatible

Methods inherited from class java.lang.Object

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

Methods inherited from interface com.yahoo.tensor.Tensor

add, approxEqual, argmax, argmin, asDouble, atan2, avg, avg, avg, bit, cellCast, concat, concat, count, count, count, divide, equal, expand, fmod, hamming, hashCode, isEmpty, join, l1Normalize, l2Normalize, larger, largerOrEqual, largest, ldexp, map, matmul, max, max, max, max, median, median, median, merge, min, min, min, min, modify, multiply, notEqual, pow, prod, prod, prod, reduce, reduce, rename, rename, size, smaller, smallerOrEqual, smallest, softmax, subtract, sum, sum, sum, toAbbreviatedString, xwPlusB

Method Details

cellIterator

public Iterator<Tensor.Cell> cellIterator()

Returns an iterator over the cells of this in the standard value order.

Specified by:

cellIterator in interface Tensor

cellIterator

public IndexedTensor.SubspaceIterator cellIterator(PartialAddress partialAddress,
 DimensionSizes iterationSizes)

Returns an iterator over all the cells in this tensor which matches the given partial address

valueIterator

public Iterator<Double> valueIterator()

Returns an iterator over the values of this returned in the standard value order

Specified by:

valueIterator in interface Tensor

subspaceIterator

public Iterator<IndexedTensor.SubspaceIterator> subspaceIterator(Set<String> dimensions,
 DimensionSizes sizes)

Returns an iterator over value iterators where the outer iterator is over each unique value of the dimensions given and the inner iterator is over each unique value of the rest of the dimensions, in the standard value order

Parameters:

dimensions - the names of the dimensions of the superspace

sizes - the size of each dimension in the space we are returning values for, containing one value per dimension of this tensor (in order). Each size may be the same or smaller than the corresponding size of this tensor

subspaceIterator

public Iterator<IndexedTensor.SubspaceIterator> subspaceIterator(Set<String> dimensions)

Returns a subspace iterator having the sizes of the dimensions of this tensor

get

public double get(long... indexes)

Returns the value at the given indexes as a double

Parameters:

indexes - the indexes into the dimensions of this. Must be one number per dimension of this

Throws:

IllegalArgumentException - if any of the indexes are out of bound or a wrong number of indexes are given

getFloat

public float getFloat(long... indexes)

Returns the value at the given indexes as a float

Parameters:

indexes - the indexes into the dimensions of this. Must be one number per dimension of this

Throws:

IllegalArgumentException - if any of the indexes are out of bound or a wrong number of indexes are given

get

public double get(TensorAddress address)

Returns the value at this address, or 0.0 if there is no value at this address

Specified by:

get in interface Tensor

has

public boolean has(TensorAddress address)

Description copied from interface: Tensor

Returns true if this cell exists

Specified by:

has in interface Tensor

get

public abstract double get(long valueIndex)

Returns the value at the given standard value order index as a double.

Parameters:

valueIndex - the direct index into the underlying data.

Throws:

IllegalArgumentException - if index is out of bounds

getFloat

public abstract float getFloat(long valueIndex)

Returns the value at the given standard value order index as a float.

Parameters:

valueIndex - the direct index into the underlying data.

Throws:

IllegalArgumentException - if index is out of bounds

type

public TensorType type()

Specified by:

type in interface Tensor

withType

public abstract IndexedTensor withType(TensorType type)

Description copied from interface: Tensor

Returns this tensor with the given type if types are compatible

Specified by:

withType in interface Tensor

dimensionSizes

public DimensionSizes dimensionSizes()

shape

public long[] shape()

cells

public Map<TensorAddress,Double> cells()

Description copied from interface: Tensor

Returns an immutable map of the cells of this in no particular order. This may be expensive for some implementations - avoid when possible

Specified by:

cells in interface Tensor

remove

public Tensor remove(Set<TensorAddress> addresses)

Description copied from interface: Tensor

Returns a new tensor where existing cells in this tensor have been removed according to the given set of addresses. Only valid for sparse or mixed tensors. For mixed tensors, addresses are assumed to only contain the sparse dimensions, as the entire dense subspace is removed.

Specified by:

remove in interface Tensor

Parameters:

addresses - list of addresses to remove

Returns:

a new tensor where cells have been removed

toString

public String toString()

Description copied from interface: Tensor

Returns this tensor on the tensor literal form with type included.

Specified by:

toString in interface Tensor

Overrides:

toString in class Object

toString

public String toString(boolean withType,
 boolean shortForms)

Description copied from interface: Tensor

Returns this tensor on the tensor literal form.

Specified by:

toString in interface Tensor

Parameters:

withType - whether to prefix the value by the type of this

shortForms - whether to use short forms where applicable, or always using the verbose form

toAbbreviatedString

public String toAbbreviatedString(boolean withType,
 boolean shortForms)

Description copied from interface: Tensor

Returns an abbreviated string representation of this tensor suitable for human-readable messages

Specified by:

toAbbreviatedString in interface Tensor

Parameters:

withType - whether to prefix the value by the type of this

shortForms - whether to use short forms where applicable, or always using the verbose form

equals

public boolean equals(Object other)

Description copied from interface: Tensor

Returns whether this tensor and the given tensor is mathematically equal: That they have the same dimension *names* and the same content.

Specified by:

equals in interface Tensor

Overrides:

equals in class Object