mikera.arrayz

Class NDArray

java.lang.Object

mikera.arrayz.impl.AbstractArray<INDArray>

mikera.arrayz.impl.BaseShapedArray

mikera.arrayz.impl.BaseNDArray

mikera.arrayz.NDArray

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable, java.lang.Comparable<INDArray>, java.lang.Iterable<INDArray>, IStridedArray, INDArray

public final class NDArray
extends BaseNDArray

General purpose dense strided NDArray class. Allows arbitrary strided access over a dense double[] array.

See Also:

Serialized Form

Method Summary

Modifier and Type	Method and Description
void	applyOp(IOperator op) Applies a unary operator to all elements of the array (in-place)
void	applyOp(Op op) Applies a unary operator to all elements of the array (in-place)
double[]	asDoubleArray() Returns the underlying double array representing the densely packed elements of this array.
AVector	asVector() Constructs a view of all elements in the array as a single vector in row-major order.
INDArray	broadcast(int... dimensions) Returns a view of this array broadcasted up to a larger shape
INDArray	clone() Returns a clone of the array data, as a new array which will be fully mutable and may be of a different class to the original.
java.util.Iterator<java.lang.Double>	elementIterator() Returns an iterator over all elements in this array, in row-major order
boolean	equals(INDArray a) Checks if two arrays are equal exactly in terms of both value and shape Element equality checks are consistent with compareTo
NDArray	exactClone() Creates a deep clone of an array, using the same class implementation as the original array
double[]	getArray() Gets the underlying data array for this strided array Unsafe operation - may allow modification of otherwise immutable arrays
java.util.List<INDArray>	getSlices() Returns a list of all major slices of this array.
INDArray	getTranspose() Returns the transpose of this array.
INDArray	getTransposeView() Returns a transposed view of the array.
INDArray	immutable() Returns an immutable version of this array.
boolean	isElementConstrained() Returns true if this array has some constraints on element values
boolean	isFullyMutable() If this method returns true, the INDArray is guaranteed to be fully mutable in all positions i.e.
boolean	isMutable() Returns true if the INDArray is mutable (at least partially)
boolean	isView() Return true if this array is considered as a view type.
void	multiply(double d) Multiplies all elements of the array in place by a given double value
static NDArray	newArray(int... shape) Creates a new zero-filled NDArray with the given shape.
INDArray	reshape(int... dimensions) Returns a new array by rearranging the elements of this array into the desired shape Truncates or zero-pads the elements as required to fill the new shape
void	set(double value) Sets all elements of an array to a specific double value
void	set(int[] indexes, double value) Sets the element at the given indexed position in a mutable array
void	set(int x, double value) Sets a value at a given position in a mutable 1D array
void	set(int x, int y, double value) Sets a value at a given position in a mutable 2D array
void	setElements(int pos, double[] values, int offset, int length) Sets elements in an array using the given double values.
INDArray	slice(int majorSlice) Returns the specified major slice of this array as a view (slice along dimension 0) 0 dimensional slice results are permitted, but attempting to slice a 0 dimensional array will result in an error.
INDArray	slice(int dimension, int index) Returns a slice view of this array along the specified dimension
NDArray	subArray(int[] offsets, int[] shape) Returns a subarray view of a larger array
void	toDoubleBuffer(java.nio.DoubleBuffer dest) Copies the elements of this INDArray to the specified double buffer.
void	validate() Validates the internal data structure of the INDArray.
static NDArray	wrap(double[] data, int[] shape) Wraps the given double[] data in an NDArray with the given shape.
static NDArray	wrap(INDArray a)
static NDArray	wrap(IStridedArray a)
static NDArray	wrap(Matrix m)
static NDArray	wrap(Vector v)
static INDArray	wrapStrided(double[] data, int offset, int[] shape, int[] strides)

Methods inherited from class mikera.arrayz.impl.BaseNDArray

dimensionality, elementCount, get, get, get, get, getArrayOffset, getIndex, getLongShape, getShape, getStride, getStrides, isPackedArray

Methods inherited from class mikera.arrayz.impl.BaseShapedArray

getElement, getShape, getShapeClone, isZero, reduce, reduce, sliceCount

Methods inherited from class mikera.arrayz.impl.AbstractArray

abs, absCopy, absDiff, absDiffCopy, add, add, add, addApplyOp, addAt, addCopy, addCopy, addCopy, addInnerProduct, addInnerProduct, addInnerProduct, addInnerProduct, addMultiple, addMultipleSparse, addOuterProduct, addOuterProductSparse, addPower, addPower, addSparse, addSparse, addToArray, applyOp, applyOp, applyOpCopy, asElementList, broadcastCloneLike, broadcastCopyLike, broadcastLike, broadcastLike, broadcastLike, clamp, compareTo, componentCount, copy, dense, denseClone, density, divide, divide, divideCopy, divideCopy, elementAbsPowSum, elementMax, elementMaxAbs, elementMin, elementPowSum, elementProduct, elementsEqual, elementSquaredSum, elementSum, ensureMutable, epsilonEquals, epsilonEquals, equals, equals, equalsArray, equalsArray, exp, fill, get, get, get, get, get, getComponent, getComponents, getElements, getElements, getElements, getElements, getSlices, getSliceViews, getTransposeCopy, hashCode, hasUncountable, innerProduct, innerProduct, innerProduct, innerProduct, isBoolean, isDense, isSameShape, isSparse, iterator, join, join, log, multiply, multiplyCopy, multiplyCopy, mutable, negate, negateCopy, nonZeroCount, outerProduct, pow, reciprocal, reciprocalCopy, reduceSlices, reduceSlices, reorder, reorder, rotateView, scale, scaleAdd, scaleAdd, scaleCopy, set, set, set, setApplyOp, setElements, setElements, setInnerProduct, setMultiple, setMultiple, setSparse, setSparse, signum, signumCopy, sliceValue, sparse, sparseClone, sqrt, square, squareCopy, sub, sub, sub, subCopy, toDoubleArray, toSliceArray, toString, toStringFull, toVector, withComponents

Methods inherited from class java.lang.Object

getClass, notify, notifyAll, wait, wait, wait

Methods inherited from interface mikera.arrayz.INDArray

abs, absCopy, absDiff, absDiffCopy, add, add, addApplyOp, addAt, addCopy, addCopy, addInnerProduct, addInnerProduct, addMultiple, addMultipleSparse, addOuterProduct, addOuterProductSparse, addPower, addPower, addSparse, addSparse, addToArray, applyOp, applyOp, applyOpCopy, asElementList, broadcastCloneLike, broadcastCopyLike, broadcastLike, broadcastLike, broadcastLike, clamp, componentCount, copy, dense, denseClone, divide, divide, divideCopy, divideCopy, elementAbsPowSum, elementMax, elementMaxAbs, elementMin, elementPowSum, elementProduct, elementsEqual, elementSquaredSum, elementSum, ensureMutable, epsilonEquals, epsilonEquals, equalsArray, equalsArray, exp, fill, get, get, get, get, get, getComponent, getComponents, getElement, getElements, getElements, getElements, getElements, getShape, getShapeClone, getSlices, getSliceViews, getTransposeCopy, hasUncountable, innerProduct, innerProduct, innerProduct, innerProduct, isBoolean, isDense, isSameShape, isSparse, isZero, join, join, log, multiply, multiplyCopy, multiplyCopy, mutable, negate, negateCopy, nonZeroCount, outerProduct, pow, reciprocal, reciprocalCopy, reduce, reduce, reduceSlices, reduceSlices, reorder, reorder, rotateView, scale, scaleAdd, scaleAdd, scaleCopy, set, set, set, setApplyOp, setElements, setElements, setInnerProduct, setMultiple, setMultiple, setSparse, setSparse, signum, signumCopy, sliceCount, sliceValue, sparse, sparseClone, sqrt, square, squareCopy, sub, sub, subCopy, toDoubleArray, toSliceArray, toVector, withComponents

Methods inherited from interface java.lang.Comparable

compareTo

Methods inherited from interface java.lang.Iterable

forEach, spliterator

Method Detail

wrap

public static NDArray wrap(double[] data,
                           int[] shape)

Wraps the given double[] data in an NDArray with the given shape. Does *not* take a defensive copy of either the data or shape array.

Parameters:

data -

shape -

Returns:

wrap

public static NDArray wrap(Vector v)

wrap

public static NDArray wrap(Matrix m)

wrap

public static NDArray wrap(IStridedArray a)

wrap

public static NDArray wrap(INDArray a)

newArray

public static NDArray newArray(int... shape)

Creates a new zero-filled NDArray with the given shape.

Parameters:

shape -

Returns:

set

public void set(double value)

Description copied from interface: INDArray

Sets all elements of an array to a specific double value

Specified by:

set in interface INDArray

Overrides:

set in class AbstractArray<INDArray>

set

public void set(int x,
                double value)

Description copied from interface: INDArray

Sets a value at a given position in a mutable 1D array

Specified by:

set in interface INDArray

Overrides:

set in class AbstractArray<INDArray>

set

public void set(int x,
                int y,
                double value)

Description copied from interface: INDArray

Sets a value at a given position in a mutable 2D array

Specified by:

set in interface INDArray

Overrides:

set in class AbstractArray<INDArray>

set

public void set(int[] indexes,
                double value)

Description copied from interface: INDArray

Sets the element at the given indexed position in a mutable array

getTranspose

public INDArray getTranspose()

Description copied from interface: INDArray

Returns the transpose of this array. A transpose of an array is equivalent to reversing the order of dimensions. May or may not return a view depending on the array type.

Specified by:

getTranspose in interface INDArray

Overrides:

getTranspose in class AbstractArray<INDArray>

getTransposeView

public INDArray getTransposeView()

Description copied from interface: INDArray

Returns a transposed view of the array. May throw UnsupportedOperationException if the array type does not support this capability

Specified by:

getTransposeView in interface INDArray

Overrides:

getTransposeView in class AbstractArray<INDArray>

asVector

public AVector asVector()

Description copied from interface: INDArray

Constructs a view of all elements in the array as a single vector in row-major order.

Specified by:

asVector in interface INDArray

Overrides:

asVector in class AbstractArray<INDArray>

reshape

public INDArray reshape(int... dimensions)

Description copied from interface: INDArray

Returns a new array by rearranging the elements of this array into the desired shape Truncates or zero-pads the elements as required to fill the new shape

Specified by:

reshape in interface INDArray

Overrides:

reshape in class AbstractArray<INDArray>

broadcast

public INDArray broadcast(int... dimensions)

Description copied from interface: INDArray

Returns a view of this array broadcasted up to a larger shape

Specified by:

broadcast in interface INDArray

Overrides:

broadcast in class AbstractArray<INDArray>

slice

public INDArray slice(int majorSlice)

Description copied from interface: INDArray

Returns the specified major slice of this array as a view (slice along dimension 0) 0 dimensional slice results are permitted, but attempting to slice a 0 dimensional array will result in an error.

slice

public INDArray slice(int dimension,
                      int index)

Description copied from interface: INDArray

Returns a slice view of this array along the specified dimension

subArray

public NDArray subArray(int[] offsets,
                        int[] shape)

Description copied from interface: INDArray

Returns a subarray view of a larger array

Specified by:

subArray in interface INDArray

Overrides:

subArray in class AbstractArray<INDArray>

isMutable

public boolean isMutable()

Description copied from interface: INDArray

Returns true if the INDArray is mutable (at least partially)

Specified by:

isMutable in interface INDArray

Overrides:

isMutable in class AbstractArray<INDArray>

isFullyMutable

public boolean isFullyMutable()

Description copied from interface: INDArray

If this method returns true, the INDArray is guaranteed to be fully mutable in all positions i.e. every position can store any valid double value

Specified by:

isFullyMutable in interface INDArray

Overrides:

isFullyMutable in class AbstractArray<INDArray>

isElementConstrained

public boolean isElementConstrained()

Description copied from interface: INDArray

Returns true if this array has some constraints on element values

Specified by:

isElementConstrained in interface INDArray

Overrides:

isElementConstrained in class AbstractArray<INDArray>

isView

public boolean isView()

Description copied from interface: INDArray

Return true if this array is considered as a view type. This indicates (but does not guarantee): a) Data is *likely* to be shared with other arrays, so mutation of one may affect others b) It is *unlikely* to be in the most efficient general purpose format isView is intended to be used as for heuristics. It should *not* be used where the outcome might affect correctness.

applyOp

public void applyOp(Op op)

Description copied from interface: INDArray

Applies a unary operator to all elements of the array (in-place)

Specified by:

applyOp in interface INDArray

Overrides:

applyOp in class AbstractArray<INDArray>

applyOp

public void applyOp(IOperator op)

Description copied from interface: INDArray

Applies a unary operator to all elements of the array (in-place)

Specified by:

applyOp in interface INDArray

Overrides:

applyOp in class AbstractArray<INDArray>

Parameters:

op - The operator to apply to the array

equals

public boolean equals(INDArray a)

Description copied from interface: INDArray

Checks if two arrays are equal exactly in terms of both value and shape Element equality checks are consistent with compareTo

Specified by:

equals in interface INDArray

Overrides:

equals in class AbstractArray<INDArray>

exactClone

public NDArray exactClone()

Description copied from interface: INDArray

Creates a deep clone of an array, using the same class implementation as the original array

Returns:

A clone of the original array

clone

public INDArray clone()

Description copied from interface: INDArray

Returns a clone of the array data, as a new array which will be fully mutable and may be of a different class to the original. Clone should attempt to return the most efficient possible array type. Clone should preserve sparsity property where possible, but this is not guaranteed.

Specified by:

clone in interface INDArray

Overrides:

clone in class AbstractArray<INDArray>

multiply

public void multiply(double d)

Description copied from interface: INDArray

Multiplies all elements of the array in place by a given double value

Specified by:

multiply in interface INDArray

Overrides:

multiply in class AbstractArray<INDArray>

setElements

public void setElements(int pos,
                        double[] values,
                        int offset,
                        int length)

Description copied from interface: INDArray

Sets elements in an array using the given double values. The source array must contain at least the specified length in terms of number of elements

Specified by:

setElements in interface INDArray

Overrides:

setElements in class AbstractArray<INDArray>

Parameters:

pos - Offset into this array, expressed in terms of number of elements in row major order

values - Element values in a double[] array. There must be at least as many element values as elements in this array.

offset - Position in the values array from which to start taking values.

length - Number of elements to set.

toDoubleBuffer

public void toDoubleBuffer(java.nio.DoubleBuffer dest)

Description copied from interface: INDArray

Copies the elements of this INDArray to the specified double buffer.

Specified by:

toDoubleBuffer in interface INDArray

Overrides:

toDoubleBuffer in class AbstractArray<INDArray>

asDoubleArray

public double[] asDoubleArray()

Description copied from interface: INDArray

Returns the underlying double array representing the densely packed elements of this array. Modifications to this double array will change the original array. Returns null if there is no such array.

Specified by:

asDoubleArray in interface INDArray

Overrides:

asDoubleArray in class AbstractArray<INDArray>

getSlices

public java.util.List<INDArray> getSlices()

Description copied from interface: INDArray

Returns a list of all major slices of this array. Returns a list of Double values if a 1-dimensional array is sliced, otherwise a list of INDArray instances

Specified by:

getSlices in interface INDArray

Overrides:

getSlices in class AbstractArray<INDArray>

elementIterator

public java.util.Iterator<java.lang.Double> elementIterator()

Description copied from interface: INDArray

Returns an iterator over all elements in this array, in row-major order

Specified by:

elementIterator in interface INDArray

Overrides:

elementIterator in class AbstractArray<INDArray>

validate

public void validate()

Description copied from interface: INDArray

Validates the internal data structure of the INDArray. Throws an exception on failure. Failure indicates a serious bug and/or data corruption.

Specified by:

validate in interface INDArray

Overrides:

validate in class AbstractArray<INDArray>

immutable

public INDArray immutable()

Description copied from interface: INDArray

Returns an immutable version of this array. May return the same array if already immutable. Guarantees a defensive copy if the array is mutable.

Specified by:

immutable in interface INDArray

Overrides:

immutable in class AbstractArray<INDArray>

getArray

public double[] getArray()

Description copied from interface: IStridedArray

Gets the underlying data array for this strided array Unsafe operation - may allow modification of otherwise immutable arrays

Returns:

wrapStrided

public static INDArray wrapStrided(double[] data,
                                   int offset,
                                   int[] shape,
                                   int[] strides)