io.improbable.keanu.tensor.dbl

Class Nd4jDoubleTensor

java.lang.Object

io.improbable.keanu.tensor.ndj4.Nd4jTensor<T,TENSOR>

io.improbable.keanu.tensor.ndj4.Nd4jNumberTensor<T,TENSOR>

io.improbable.keanu.tensor.ndj4.Nd4jFloatingPointTensor<java.lang.Double,DoubleTensor>

io.improbable.keanu.tensor.dbl.Nd4jDoubleTensor

All Implemented Interfaces:

io.improbable.keanu.kotlin.DoubleOperators<DoubleTensor>, io.improbable.keanu.kotlin.NumberOperators<DoubleTensor>, DoubleTensor, FloatingPointTensor<java.lang.Double,DoubleTensor>, NumberTensor<java.lang.Double,DoubleTensor>, Tensor<java.lang.Double,DoubleTensor>

public class Nd4jDoubleTensor
extends Nd4jFloatingPointTensor<java.lang.Double,DoubleTensor>
implements DoubleTensor

Class for representing n-dimensional arrays of doubles. This is backed by Nd4j.

Nested Class Summary

Nested classes/interfaces inherited from interface io.improbable.keanu.tensor.Tensor

Tensor.FlattenedView<N>

Field Summary

Fields inherited from class io.improbable.keanu.tensor.ndj4.Nd4jTensor

tensor

Fields inherited from interface io.improbable.keanu.tensor.Tensor

ONE_BY_ONE_SHAPE, SCALAR_SHAPE, SCALAR_STRIDE

Constructor Summary

Constructors

Constructor and Description
Nd4jDoubleTensor(double[] data, long[] shape)
Nd4jDoubleTensor(org.nd4j.linalg.api.ndarray.INDArray tensor)

Method Summary

Modifier and Type	Method and Description
static Nd4jDoubleTensor	arange(double start, double end)
static Nd4jDoubleTensor	arange(double start, double end, double stepSize)
int	argMax() Find the index into the flattened array of the tensor of the largest value, e.g.
IntegerTensor	argMax(int axis) Find the indices into the tensor of the largest values in a specified axis (dimension), e.g.
int	argMin()
IntegerTensor	argMin(int axis)
java.lang.Double[]	asFlatArray()
double[]	asFlatDoubleArray()
int[]	asFlatIntegerArray()
static Nd4jDoubleTensor	create(double[] values)
static Nd4jDoubleTensor	create(double[] values, long... shape)
static Nd4jDoubleTensor	create(double value, long... shape)
protected DoubleTensor	create(org.nd4j.linalg.api.ndarray.INDArray tensor)
DoubleTensor	digammaInPlace()
boolean	equalsWithinEpsilon(DoubleTensor o, java.lang.Double epsilon)
static Nd4jDoubleTensor	eye(long n)
Tensor.FlattenedView<java.lang.Double>	getFlattenedView()
protected java.lang.Double	getNumber(java.lang.Number number)
protected org.nd4j.linalg.api.ndarray.INDArray	getTensor(Tensor tensor)
protected DoubleTensor	getThis()
DoubleTensor	greaterThanMask(DoubleTensor greaterThanThis)
DoubleTensor	greaterThanOrEqualToMask(DoubleTensor greaterThanOrEqualToThis)
BooleanTensor	isFinite()
BooleanTensor	isInfinite()
BooleanTensor	isNegativeInfinity()
BooleanTensor	isPositiveInfinity()
DoubleTensor	lessThanMask(DoubleTensor lessThanThis)
DoubleTensor	lessThanOrEqualToMask(DoubleTensor lessThanOrEqualToThis)
static Nd4jDoubleTensor	linspace(double start, double end, int numberOfPoints)
DoubleTensor	logAddExp2InPlace(DoubleTensor that)
DoubleTensor	logAddExpInPlace(DoubleTensor that)
DoubleTensor	logGammaInPlace()
int	nanArgMax()
IntegerTensor	nanArgMax(int axis)
int	nanArgMin()
IntegerTensor	nanArgMin(int axis)
static Nd4jDoubleTensor	ones(long... shape)
DoubleTensor	replaceNaNInPlace(java.lang.Double value)
DoubleTensor	safeLogTimesInPlace(DoubleTensor y)
static Nd4jDoubleTensor	scalar(double scalarValue)
protected DoubleTensor	set(org.nd4j.linalg.api.ndarray.INDArray tensor)
DoubleTensor	toDouble()
IntegerTensor	toInteger()
static Nd4jDoubleTensor	zeros(long[] shape)

Methods inherited from class io.improbable.keanu.tensor.ndj4.Nd4jFloatingPointTensor

acoshInPlace, acosInPlace, asinhInPlace, asinInPlace, atan2InPlace, atan2InPlace, atanhInPlace, atanInPlace, ceilInPlace, choleskyDecomposition, clampInPlace, coshInPlace, cosInPlace, determinant, exp2InPlace, expInPlace, expM1InPlace, floorInPlace, log10InPlace, log1pInPlace, log2InPlace, logInPlace, matrixInverse, matrixMultiply, notNaN, product, reciprocalInPlace, roundInPlace, setAllInPlace, sigmoidInPlace, sinhInPlace, sinInPlace, sqrtInPlace, standardizeInPlace, tanhInPlace, tanInPlace, tensorMultiply

Methods inherited from class io.improbable.keanu.tensor.ndj4.Nd4jNumberTensor

absInPlace, applyInPlace, average, cumProdInPlace, cumSumInPlace, div, divInPlace, divInPlace, elementwiseEquals, elementwiseEquals, fromMask, greaterThan, greaterThan, greaterThanOrEqual, greaterThanOrEqual, lessThan, lessThan, lessThanOrEqual, lessThanOrEqual, max, maxInPlace, min, minInPlace, minus, minusInPlace, minusInPlace, plus, plusInPlace, plusInPlace, powInPlace, powInPlace, product, reverseDivInPlace, reverseDivInPlace, reverseMinusInPlace, reverseMinusInPlace, setWithMaskInPlace, standardDeviation, sum, sum, times, timesInPlace, timesInPlace, unaryMinusInPlace

Methods inherited from class io.improbable.keanu.tensor.ndj4.Nd4jTensor

broadcast, diag, duplicate, equals, get, getLength, getRank, getShape, getStride, getTensor, hashCode, permute, reshape, slice, slice, split, take, toString, transpose

Methods inherited from class java.lang.Object

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

Methods inherited from interface io.improbable.keanu.tensor.dbl.DoubleTensor

arange, concat, concat, div, max, min, minus, plus, pow, reverseDiv, reverseMinus, stack, times, vector

Methods inherited from interface io.improbable.keanu.tensor.FloatingPointTensor

acos, acosh, acoshInPlace, acosInPlace, asin, asinh, asinhInPlace, asinInPlace, atan, atan2, atan2, atan2InPlace, atan2InPlace, atanh, atanhInPlace, atanInPlace, ceil, ceilInPlace, choleskyDecomposition, cos, cosh, coshInPlace, cosInPlace, determinant, digamma, exp, exp2, exp2InPlace, expInPlace, expM1, expM1InPlace, floor, floorInPlace, isNaN, log, log10, log10InPlace, log1p, log1pInPlace, log2, log2InPlace, logAddExp, logAddExp2, logGamma, logInPlace, matrixInverse, notNaN, reciprocal, reciprocalInPlace, replaceNaN, round, roundInPlace, sigmoid, sigmoidInPlace, sin, sinh, sinhInPlace, sinInPlace, sqrt, sqrtInPlace, standardize, standardizeInPlace, tan, tanh, tanhInPlace, tanInPlace

Methods inherited from interface io.improbable.keanu.tensor.NumberTensor

abs, absInPlace, apply, applyInPlace, average, clamp, clampInPlace, cumProd, cumProdInPlace, cumSum, cumSumInPlace, div, div, divInPlace, divInPlace, greaterThan, greaterThan, greaterThanOrEqual, greaterThanOrEqual, lessThan, lessThan, lessThanOrEqual, lessThanOrEqual, matrixMultiply, max, max, maxInPlace, min, min, minInPlace, minus, minus, minusInPlace, minusInPlace, plus, plus, plusInPlace, plusInPlace, pow, pow, powInPlace, powInPlace, product, product, reverseDiv, reverseDiv, reverseDivInPlace, reverseDivInPlace, reverseMinus, reverseMinus, reverseMinusInPlace, reverseMinusInPlace, safeLogTimes, setAllInPlace, setWithMask, setWithMaskInPlace, standardDeviation, sum, sum, tensorMultiply, times, times, timesInPlace, timesInPlace, unaryMinus, unaryMinusInPlace

Methods inherited from interface io.improbable.keanu.tensor.Tensor

asFlatList, broadcast, create, createFilled, diag, duplicate, elementwiseEquals, elementwiseEquals, elementwiseEquals, expandDims, get, getLength, getRank, getShape, getStride, getValue, hasSameShapeAs, hasSameShapeAs, isLengthOne, isMatrix, isScalar, isVector, moveAxis, permute, reshape, scalar, setValue, slice, slice, slice, sliceAlongDimension, split, squeeze, swapAxis, take, transpose

Constructor Detail

Nd4jDoubleTensor

public Nd4jDoubleTensor(double[] data,
                        long[] shape)

Nd4jDoubleTensor

public Nd4jDoubleTensor(org.nd4j.linalg.api.ndarray.INDArray tensor)

Method Detail

getTensor

protected org.nd4j.linalg.api.ndarray.INDArray getTensor(Tensor tensor)

Specified by:

getTensor in class Nd4jTensor<java.lang.Double,DoubleTensor>

create

protected DoubleTensor create(org.nd4j.linalg.api.ndarray.INDArray tensor)

Specified by:

create in class Nd4jTensor<java.lang.Double,DoubleTensor>

set

protected DoubleTensor set(org.nd4j.linalg.api.ndarray.INDArray tensor)

Specified by:

set in class Nd4jTensor<java.lang.Double,DoubleTensor>

getThis

protected DoubleTensor getThis()

Specified by:

getThis in class Nd4jTensor<java.lang.Double,DoubleTensor>

scalar

public static Nd4jDoubleTensor scalar(double scalarValue)

Specified by:

scalar in interface DoubleTensor

Specified by:

scalar in interface Tensor<java.lang.Double,DoubleTensor>

create

public static Nd4jDoubleTensor create(double[] values,
                                      long... shape)

Specified by:

create in interface DoubleTensor

create

public static Nd4jDoubleTensor create(double value,
                                      long... shape)

Specified by:

create in interface DoubleTensor

create

public static Nd4jDoubleTensor create(double[] values)

Specified by:

create in interface DoubleTensor

ones

public static Nd4jDoubleTensor ones(long... shape)

Specified by:

ones in interface DoubleTensor

eye

public static Nd4jDoubleTensor eye(long n)

Specified by:

eye in interface DoubleTensor

zeros

public static Nd4jDoubleTensor zeros(long[] shape)

Specified by:

zeros in interface DoubleTensor

linspace

public static Nd4jDoubleTensor linspace(double start,
                                        double end,
                                        int numberOfPoints)

Specified by:

linspace in interface DoubleTensor

arange

public static Nd4jDoubleTensor arange(double start,
                                      double end)

Specified by:

arange in interface DoubleTensor

Parameters:

start - start of range

end - end of range (exclusive)

Returns:

a vector of numbers from start incrementing by one to end (exclusively)

arange

public static Nd4jDoubleTensor arange(double start,
                                      double end,
                                      double stepSize)

Specified by:

arange in interface DoubleTensor

Parameters:

start - start of range

end - end of range (exclusive)

stepSize - size of step from start to end

Returns:

a vector of numbers starting at start and stepping to end (exclusively)

nanArgMax

public int nanArgMax()

Specified by:

nanArgMax in interface FloatingPointTensor<java.lang.Double,DoubleTensor>

argMax

public int argMax()

Description copied from interface: NumberTensor

Find the index into the flattened array of the tensor of the largest value, e.g.

 DoubleTensor tensor = DoubleTensor.arange(0, 6).reshape(2, 3);
 // [[0., 1., 2.],
 //  [3., 4., 5.]]
 IntegerTensor max = tensor.argMax();
 // [[5]]
 

Specified by:

argMax in interface NumberTensor<java.lang.Double,DoubleTensor>

Overrides:

argMax in class Nd4jNumberTensor<java.lang.Double,DoubleTensor>

Returns:

A scalar tensor with the index of the largest value. If there are multiple largest values, it will be the first index.

nanArgMax

public IntegerTensor nanArgMax(int axis)

Specified by:

nanArgMax in interface FloatingPointTensor<java.lang.Double,DoubleTensor>

argMax

public IntegerTensor argMax(int axis)

Description copied from interface: NumberTensor

Find the indices into the tensor of the largest values in a specified axis (dimension), e.g.

 DoubleTensor tensor = DoubleTensor.arange(0, 6).reshape(2, 3);
 // [[0., 1., 2.],
 //  [3., 4., 5.]]
 IntegerTensor maxesFor0 = tensor.argMax(0);
 // [[1, 1, 1]]
 IntegerTensor maxFor1 = tensor.argMax(1);
 // [[2, 2]]
 

Specified by:

argMax in interface NumberTensor<java.lang.Double,DoubleTensor>

Overrides:

argMax in class Nd4jNumberTensor<java.lang.Double,DoubleTensor>

Parameters:

axis - The axis (dimension) to find the largest values in

Returns:

A tensor where each value is the location of the maximum value in the vector at that location in the specified dimension in the original tensor. If there are multiple largest values in this vector, it will be the first index.

See Also:

An article about argmax over an axis in numpy

nanArgMin

public int nanArgMin()

Specified by:

nanArgMin in interface FloatingPointTensor<java.lang.Double,DoubleTensor>

argMin

public int argMin()

Specified by:

argMin in interface NumberTensor<java.lang.Double,DoubleTensor>

Overrides:

argMin in class Nd4jNumberTensor<java.lang.Double,DoubleTensor>

nanArgMin

public IntegerTensor nanArgMin(int axis)

Specified by:

nanArgMin in interface FloatingPointTensor<java.lang.Double,DoubleTensor>

argMin

public IntegerTensor argMin(int axis)

Specified by:

argMin in interface NumberTensor<java.lang.Double,DoubleTensor>

Overrides:

argMin in class Nd4jNumberTensor<java.lang.Double,DoubleTensor>

getNumber

protected java.lang.Double getNumber(java.lang.Number number)

Specified by:

getNumber in class Nd4jNumberTensor<java.lang.Double,DoubleTensor>

equalsWithinEpsilon

public boolean equalsWithinEpsilon(DoubleTensor o,
                                   java.lang.Double epsilon)

Specified by:

equalsWithinEpsilon in interface NumberTensor<java.lang.Double,DoubleTensor>

greaterThanMask

public DoubleTensor greaterThanMask(DoubleTensor greaterThanThis)

Specified by:

greaterThanMask in interface NumberTensor<java.lang.Double,DoubleTensor>

greaterThanOrEqualToMask

public DoubleTensor greaterThanOrEqualToMask(DoubleTensor greaterThanOrEqualToThis)

Specified by:

greaterThanOrEqualToMask in interface NumberTensor<java.lang.Double,DoubleTensor>

lessThanMask

public DoubleTensor lessThanMask(DoubleTensor lessThanThis)

Specified by:

lessThanMask in interface NumberTensor<java.lang.Double,DoubleTensor>

lessThanOrEqualToMask

public DoubleTensor lessThanOrEqualToMask(DoubleTensor lessThanOrEqualToThis)

Specified by:

lessThanOrEqualToMask in interface NumberTensor<java.lang.Double,DoubleTensor>

safeLogTimesInPlace

public DoubleTensor safeLogTimesInPlace(DoubleTensor y)

Specified by:

safeLogTimesInPlace in interface NumberTensor<java.lang.Double,DoubleTensor>

logGammaInPlace

public DoubleTensor logGammaInPlace()

Specified by:

logGammaInPlace in interface FloatingPointTensor<java.lang.Double,DoubleTensor>

digammaInPlace

public DoubleTensor digammaInPlace()

Specified by:

digammaInPlace in interface FloatingPointTensor<java.lang.Double,DoubleTensor>

logAddExp2InPlace

public DoubleTensor logAddExp2InPlace(DoubleTensor that)

Specified by:

logAddExp2InPlace in interface FloatingPointTensor<java.lang.Double,DoubleTensor>

logAddExpInPlace

public DoubleTensor logAddExpInPlace(DoubleTensor that)

Specified by:

logAddExpInPlace in interface FloatingPointTensor<java.lang.Double,DoubleTensor>

replaceNaNInPlace

public DoubleTensor replaceNaNInPlace(java.lang.Double value)

Specified by:

replaceNaNInPlace in interface FloatingPointTensor<java.lang.Double,DoubleTensor>

isFinite

public BooleanTensor isFinite()

Specified by:

isFinite in interface FloatingPointTensor<java.lang.Double,DoubleTensor>

isInfinite

public BooleanTensor isInfinite()

Specified by:

isInfinite in interface FloatingPointTensor<java.lang.Double,DoubleTensor>

isNegativeInfinity

public BooleanTensor isNegativeInfinity()

Specified by:

isNegativeInfinity in interface FloatingPointTensor<java.lang.Double,DoubleTensor>

isPositiveInfinity

public BooleanTensor isPositiveInfinity()

Specified by:

isPositiveInfinity in interface FloatingPointTensor<java.lang.Double,DoubleTensor>

toDouble

public DoubleTensor toDouble()

Specified by:

toDouble in interface NumberTensor<java.lang.Double,DoubleTensor>

toInteger

public IntegerTensor toInteger()

Specified by:

toInteger in interface NumberTensor<java.lang.Double,DoubleTensor>

asFlatDoubleArray

public double[] asFlatDoubleArray()

Specified by:

asFlatDoubleArray in interface NumberTensor<java.lang.Double,DoubleTensor>

asFlatIntegerArray

public int[] asFlatIntegerArray()

Specified by:

asFlatIntegerArray in interface NumberTensor<java.lang.Double,DoubleTensor>

asFlatArray

public java.lang.Double[] asFlatArray()

Specified by:

asFlatArray in interface Tensor<java.lang.Double,DoubleTensor>

getFlattenedView

public Tensor.FlattenedView<java.lang.Double> getFlattenedView()

Specified by:

getFlattenedView in interface Tensor<java.lang.Double,DoubleTensor>