neureka.ndim.config.types.permuted

Class PermutedNDConfiguration

java.lang.Object

neureka.ndim.config.AbstractNDC

neureka.ndim.config.types.permuted.PermutedNDConfiguration

All Implemented Interfaces:

NDConfiguration

public class PermutedNDConfiguration
extends AbstractNDC

Nested Class Summary

Nested classes/interfaces inherited from interface neureka.ndim.config.NDConfiguration

NDConfiguration.IndexToIndexFunction, NDConfiguration.Layout, NDConfiguration.Utility

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	PermutedNDConfiguration(int[] shape, int[] strides, int[] indicesMap)

Method Summary

Modifier and Type	Method and Description
static PermutedNDConfiguration	construct(int[] shape, int[] strides, int[] indicesMap)
int	indexOfIndex(int index) Use this to calculate the true index for an element in the data array (data array index) based on a provided "virtual index", or "value array index".
int	indexOfIndices(int[] indices) The following method calculates the true index for an element in the data array based on a provided index array.
int[]	indicesMap() If one wants to for example access the fourth last item of all items within a tensor based on a scalar index x then the NDConfiguration.indicesMap() is needed as a basis for translating said scalar index x to an array of indices for every axis of the tensor represented by this NDConfiguration.
int	indicesMap(int i) This method receives an axis index and return the indices mapping value of said axis to enable readable access to the indices map of this configuration.
int[]	indicesOfIndex(int index) The following method calculates the axis indices for an element in the nd-array array based on a provided "virtual index".
int[]	offset() The offset is the position of a slice within the n-dimensional data array of its parent tensor.
int	offset(int i) The offset is the position of a slice within the n-dimensional data array of its parent tensor.
int	rank() This method returns the number of axis of a nd-array / Tensor which is equal to the length of the shape of an nd-array / Tensor.
int[]	shape() This method returns an array of axis sizes.
int	shape(int i) This method receives an axis index and return the size of the axis.
int[]	spread() The spread is the access step size of a slice within the n-dimensional data array of its parent tensor.
int	spread(int i) The spread is the access step size of a slice within the n-dimensional data array of its parent tensor.
int[]	strides() The array returned by this method is used to translate an array of axis indices to a single ata array index.
int	strides(int i) This method receives an axis index and returns the translation value for the targeted axis.

Methods inherited from class neureka.ndim.config.AbstractNDC

_cacheArray, _cached, _simpleReshape, equals, equals, hashCode, newReshaped, toString

Methods inherited from class java.lang.Object

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

Methods inherited from interface neureka.ndim.config.NDConfiguration

asInlineArray, getIndexToIndexAccessPattern, getLayout, getTraits, has, isCompact, isSimple, isVirtual, none, of, size

Constructor Detail

PermutedNDConfiguration

protected PermutedNDConfiguration(int[] shape,
                                  int[] strides,
                                  int[] indicesMap)

Method Detail

construct

public static PermutedNDConfiguration construct(int[] shape,
                                                int[] strides,
                                                int[] indicesMap)

rank

public final int rank()

This method returns the number of axis of a nd-array / Tensor which is equal to the length of the shape of an nd-array / Tensor.

Returns:

The number of axis of an nd-array.

shape

public final int[] shape()

This method returns an array of axis sizes.

Returns:

An array of axis sizes.

shape

public final int shape(int i)

This method receives an axis index and return the size of the axis. It enables readable access to the shape of this configuration.

Parameters:

i - The index of the axis whose size ought to be returned.

Returns:

The axis size targeted by the provided index.

indicesMap

public final int[] indicesMap()

If one wants to for example access the fourth last item of all items within a tensor based on a scalar index x then the NDConfiguration.indicesMap() is needed as a basis for translating said scalar index x to an array of indices for every axis of the tensor represented by this NDConfiguration.

Returns:

An array of values which are used to map an index to an indices array.

indicesMap

public final int indicesMap(int i)

This method receives an axis index and return the indices mapping value of said axis to enable readable access to the indices map of this configuration. If one wants to for example access the fourth last item of all items within a tensor based on a scalar index x then the NDConfiguration.indicesMap() is needed as a basis for translating said scalar index x to an array of indices for every axis of the tensor represented by this NDConfiguration.

Parameters:

i - The index of the axis whose indices map value ought to be returned.

Returns:

The indices map value targeted by the provided index.

strides

public final int[] strides()

The array returned by this method is used to translate an array of axis indices to a single ata array index. It is used alongside NDConfiguration.spread() and NDConfiguration.offset() by the NDConfiguration.indexOfIndices(int[]) method.

Returns:

An array of values used to translate the axes indices to a data array index.

strides

public final int strides(int i)

This method receives an axis index and returns the translation value for the targeted axis. It enables readable and fast access to the translation of this configuration.

Parameters:

i - The index of the axis whose translation ought to be returned.

Returns:

The axis translation targeted by the provided index.

spread

public final int[] spread()

The spread is the access step size of a slice within the n-dimensional data array of its parent tensor.

Returns:

An array of index step sizes for each tensor dimension / axis.

spread

public final int spread(int i)

The spread is the access step size of a slice within the n-dimensional data array of its parent tensor. Use this to look up the spread in a particular dimension / axis.

Parameters:

i - The dimension / axis index of the dimension / axis whose spread should be returned.

Returns:

The spread of the targeted dimension.

offset

public final int[] offset()

The offset is the position of a slice within the n-dimensional data array of its parent tensor. Use this to get the offsets of all slice dimension.

Returns:

The offset position of the slice tensor inside the n-dimensional data array of the parent tensor.

offset

public final int offset(int i)

The offset is the position of a slice within the n-dimensional data array of its parent tensor. Use this to look up the offset in a particular dimension / axis.

Parameters:

i - The dimension / axis index of the dimension / axis whose offset should be returned.

Returns:

The offset of the targeted dimension.

indexOfIndex

public final int indexOfIndex(int index)

Use this to calculate the true index for an element in the data array (data array index) based on a provided "virtual index", or "value array index". This virtual index may be different from the true index depending on the type of nd-array, like for example if the nd-array is a slice of another larger nd-array, or if it is in fact a permuted version of another nd-array. This virtual index ought to be turned into an index array which defines the position for every axis. Then this indices array will be converted into the final and true index targeting an underlying item. The information needed for performing this translation is expressed by individual implementations of this NDConfiguration interface, which contain everything needed to treat a given block of data as a nd-array!

Parameters:

index - The virtual index of the tensor having this configuration.

Returns:

The true index which targets the actual data within the underlying data array of an nd-array / tensor.

indicesOfIndex

public final int[] indicesOfIndex(int index)

The following method calculates the axis indices for an element in the nd-array array based on a provided "virtual index". The resulting index defines the position of the element for every axis.

Parameters:

index - The virtual index of the tensor having this configuration.

Returns:

The position of the (virtually) targeted element represented as an array of axis indices.

indexOfIndices

public final int indexOfIndices(int[] indices)

The following method calculates the true index for an element in the data array based on a provided index array.

Parameters:

indices - The indices for every axis of a given nd-array.

Returns:

The true index targeting the underlying data array of a given nd-array.