org.nd4j.linalg.api.ndarray

Class BaseSparseNDArrayCSR

java.lang.Object

org.nd4j.linalg.api.ndarray.BaseSparseNDArray

org.nd4j.linalg.api.ndarray.BaseSparseNDArrayCSR

All Implemented Interfaces:

Serializable, INDArray, ISparseNDArray

public abstract class BaseSparseNDArrayCSR
extends BaseSparseNDArray

Author:

Audrey Loeffel

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
protected DataBuffer	columnsPointers
protected static SparseFormat	format
protected DataBuffer	pointerB
protected DataBuffer	pointerE
protected DataBuffer	values

Fields inherited from class org.nd4j.linalg.api.ndarray.BaseSparseNDArray

columns, isMatrix, isScalar, isSparse, isVector, javaShapeInformation, length, rank, rows, shape, shapeInformation, sparseInformation, stride, THRESHOLD_MEMORY_ALLOCATION

Constructor Summary

Constructors

Constructor and Description
BaseSparseNDArrayCSR(DataBuffer data, int[] columnsPointers, int[] pointerB, int[] pointerE, int[] shape)
BaseSparseNDArrayCSR(double[] data, int[] columnsPointers, int[] pointerB, int[] pointerE, int[] shape) The length of the values and columns arrays is equal to the number of non-zero elements in A.
BaseSparseNDArrayCSR(float[] data, int[] columnsPointers, int[] pointerB, int[] pointerE, int[] shape)

Method Summary

Modifier and Type	Method and Description
DataBuffer	data() Returns a linear double array representation of this ndarray
boolean	equals(Object o)
INDArray	get(INDArrayIndex... indexes) Returns a subset of this array based on the specified indexes
double[]	getColumns()
double[]	getDoubleValues()
SparseFormat	getFormat() Return the sparse format (i.e COO, CSR, ...)
DataBuffer	getPointerB()
int[]	getPointerBArray()
DataBuffer	getPointerE()
int[]	getPointerEArray()
DataBuffer	getVectorCoordinates() Return the minor pointers.
boolean	isView() Returns true if this array is a view or not
INDArray	putiColumnVector(INDArray columnVector) In place assignment of a column vector
INDArray	putiRowVector(INDArray rowVector) in place assignment of row vector, to each row of this array
INDArray	putScalar(int row, int col, double value) Insert the value at the specified indices, in a 2d (rank 2) NDArray Equivalent to INDArray.putScalar(int[], double) but avoids int[] creation
DataBuffer	shapeInfoDataBuffer() Shape info
INDArray	subArray(long[] offsets, int[] shape, int[] stride)
INDArray	subArray(ShapeOffsetResolution resolution) Sub array based on the pre calculated shape,strides, offsets
INDArray	toDense() Return a dense representation of the sparse ndarray
int	underlyingRank()

Methods inherited from class org.nd4j.linalg.api.ndarray.BaseSparseNDArray

add, add, add, add, add, add, addColumnVector, addi, addi, addi, addi, addi, addi, addiColumnVector, addiRowVector, addRowVector, amax, amaxNumber, amean, ameanNumber, amin, aminNumber, argMax, assign, assign, assignIf, broadcast, broadcast, checkDimensions, cleanup, clone, columns, cond, condi, cumsum, cumsumi, detach, dimShuffle, distance1, distance2, div, div, div, div, div, div, divColumnVector, divi, divi, divi, divi, divi, divi, diviColumnVector, diviRowVector, divRowVector, dup, dup, element, elementStride, elementWiseStride, entropy, entropyNumber, eps, eps, epsi, epsi, eq, eq, eqi, eqi, equalsWithEps, flags, fmod, fmod, fmod, fmod, fmodi, fmodi, get, get, getColumn, getColumns, getDouble, getDouble, getDouble, getDoubleUnsafe, getFloat, getFloat, getFloat, getInt, getLeadingOnes, getPercentile, getRow, getRows, getScalar, getScalar, getScalar, getTrailingOnes, getWhere, getWhere, gt, gt, gte, gtei, gti, gti, hiddenDimensions, index, init, innerMostStride, isAttached, isCleanedUp, isColumnVector, isColumnVectorOrScalar, isCompressed, isInScope, isMatrix, isRowVector, isRowVectorOrScalar, isScalar, isSparse, isSquare, isVector, isVectorOrScalar, isWrapAround, javaTensorAlongDimension, length, lengthLong, leverage, leverageOrDetach, leverageTo, leverageTo, linearIndex, linearView, linearViewColumnOrder, logEntropy, logEntropyNumber, lt, lt, lte, ltei, lti, lti, majorStride, markAsCompressed, match, match, max, maxComplex, maxNumber, mean, mean, meanComplex, meanNumber, median, medianNumber, migrate, migrate, min, minComplex, minNumber, mmul, mmul, mmuli, mmuli, mul, mul, mul, mul, mul, mul, mulColumnVector, muli, muli, muli, muli, muli, muli, muliColumnVector, muliRowVector, mulRowVector, neg, negi, neq, neq, neqi, neqi, nnz, norm1, norm1Complex, norm1Number, norm2, norm2Complex, norm2Number, normmax, normmaxComplex, normmaxNumber, offset, ordering, originalOffset, percentile, percentileNumber, permute, permutei, prod, prodComplex, prodNumber, put, put, put, put, put, put, put, put, putColumn, putRow, putScalar, putScalar, putScalar, putScalar, putScalar, putScalar, putScalar, putScalar, putScalarUnsafe, putSlice, putWhere, putWhere, putWhere, putWhereWithMask, putWhereWithMask, rank, ravel, ravel, rdiv, rdiv, rdiv, rdiv, rdiv, rdiv, rdivColumnVector, rdivi, rdivi, rdivi, rdivi, rdivi, rdivi, rdiviColumnVector, rdiviRowVector, rdivRowVector, reallocate, remainder, remainder, remainder, remainder, remainderi, remainderi, repeat, replaceWhere, repmat, resetLinearView, reshape, reshape, reshape, reshape, rows, rsub, rsub, rsub, rsub, rsub, rsub, rsubColumnVector, rsubi, rsubi, rsubi, rsubi, rsubi, rsubi, rsubiColumnVector, rsubiRowVector, rsubRowVector, scan, secondaryStride, setData, setOrder, setShape, setShapeAndStride, setShapeInformation, setStride, setWrapAround, shannonEntropy, shannonEntropyNumber, shape, shapeInfo, shapeInfoToString, shapeOf, size, slice, slice, slices, sliceVectors, sparseInfoDataBuffer, sparseOffsets, squaredDistance, std, std, stdComplex, stdNumber, stdNumber, stride, stride, strideOf, sub, sub, sub, sub, sub, sub, subColumnVector, subi, subi, subi, subi, subi, subi, subiColumnVector, subiRowVector, subRowVector, sum, sum, sumComplex, sumNumber, swapAxes, tensorAlongDimension, tensorssAlongDimension, toDoubleMatrix, toDoubleVector, toFloatMatrix, toFloatVector, toIntMatrix, toIntVector, transpose, transposei, unsafeDuplication, unsafeDuplication, var, var, varComplex, varNumber, vectorAlongDimension, vectorsAlongDimension

Methods inherited from class java.lang.Object

finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface org.nd4j.linalg.api.ndarray.INDArray

convertToDoubles, convertToFloats, mmul, mmul, mmuli, mmuli, toFlatArray

Field Detail

format

protected static final SparseFormat format

values

protected transient volatile DataBuffer values

columnsPointers

protected transient volatile DataBuffer columnsPointers

pointerB

protected transient volatile DataBuffer pointerB

pointerE

protected transient volatile DataBuffer pointerE

Constructor Detail

BaseSparseNDArrayCSR

public BaseSparseNDArrayCSR(double[] data,
                            int[] columnsPointers,
                            int[] pointerB,
                            int[] pointerE,
                            int[] shape)

The length of the values and columns arrays is equal to the number of non-zero elements in A. The length of the pointerB and pointerE arrays is equal to the number of rows in A.

Parameters:

data - a double array that contains the non-zero element of the sparse matrix A

columnsPointers - Element i of the integer array columns is the number of the column in A that contains the i-th value in the values array.

pointerB - Element j of this integer array gives the index of the element in the values array that is first non-zero element in a row j of A. Note that this index is equal to pointerB(j) - pointerB(1)+1 .

pointerE - An integer array that contains row indices, such that pointerE(j)-pointerB(1) is the index of the element in the values array that is last non-zero element in a row j of A.

shape - Shape of the matrix A

BaseSparseNDArrayCSR

public BaseSparseNDArrayCSR(float[] data,
                            int[] columnsPointers,
                            int[] pointerB,
                            int[] pointerE,
                            int[] shape)

BaseSparseNDArrayCSR

public BaseSparseNDArrayCSR(DataBuffer data,
                            int[] columnsPointers,
                            int[] pointerB,
                            int[] pointerE,
                            int[] shape)

Method Detail

putScalar

public INDArray putScalar(int row,
                          int col,
                          double value)

Description copied from interface: INDArray

Insert the value at the specified indices, in a 2d (rank 2) NDArray
Equivalent to INDArray.putScalar(int[], double) but avoids int[] creation

Specified by:

putScalar in interface INDArray

Overrides:

putScalar in class BaseSparseNDArray

Parameters:

row - Row (dimension 0) index

col - Column (dimension 1) index

value - Value to put

Returns:

This INDArray

get

public INDArray get(INDArrayIndex... indexes)

Returns a subset of this array based on the specified indexes

Specified by:

get in interface INDArray

Overrides:

get in class BaseSparseNDArray

Parameters:

indexes - the indexes in to the array

Returns:

a view of the array with the specified indices

getVectorCoordinates

public DataBuffer getVectorCoordinates()

Return the minor pointers. (columns for CSR, rows for CSC,...)

Returns:

a DataBuffer of indexes

getDoubleValues

public double[] getDoubleValues()

getColumns

public double[] getColumns()

getPointerBArray

public int[] getPointerBArray()

getPointerEArray

public int[] getPointerEArray()

getFormat

public SparseFormat getFormat()

Description copied from interface: ISparseNDArray

Return the sparse format (i.e COO, CSR, ...)

Returns:

format

See Also:

SparseFormat

getPointerB

public DataBuffer getPointerB()

getPointerE

public DataBuffer getPointerE()

data

public DataBuffer data()

Description copied from interface: INDArray

Returns a linear double array representation of this ndarray

Returns:

the linear double array representation of this ndarray

toDense

public INDArray toDense()

Description copied from interface: ISparseNDArray

Return a dense representation of the sparse ndarray

Specified by:

toDense in interface INDArray

Specified by:

toDense in interface ISparseNDArray

Overrides:

toDense in class BaseSparseNDArray

shapeInfoDataBuffer

public DataBuffer shapeInfoDataBuffer()

Description copied from interface: INDArray

Shape info

Specified by:

shapeInfoDataBuffer in interface INDArray

Overrides:

shapeInfoDataBuffer in class BaseSparseNDArray

Returns:

subArray

public INDArray subArray(ShapeOffsetResolution resolution)

Description copied from interface: INDArray

Sub array based on the pre calculated shape,strides, offsets

Specified by:

subArray in interface INDArray

Overrides:

subArray in class BaseSparseNDArray

Parameters:

resolution - the resolution to use

Returns:

the sub array based on the calculations from the resolution

subArray

public INDArray subArray(long[] offsets,
                         int[] shape,
                         int[] stride)

Specified by:

subArray in interface INDArray

Overrides:

subArray in class BaseSparseNDArray

Returns:

equals

public boolean equals(Object o)

Overrides:

equals in class BaseSparseNDArray

isView

public boolean isView()

Description copied from interface: INDArray

Returns true if this array is a view or not

Returns:

underlyingRank

public int underlyingRank()

putiColumnVector

public INDArray putiColumnVector(INDArray columnVector)

Description copied from interface: INDArray

In place assignment of a column vector

Parameters:

columnVector - the column vector to add

Returns:

the result of the addition

putiRowVector

public INDArray putiRowVector(INDArray rowVector)

Description copied from interface: INDArray

in place assignment of row vector, to each row of this array

Parameters:

rowVector - Row vector to put

Returns:

This array, after assigning every road to the specified value