Package org.apache.sysds.runtime.matrix.data.sketch.countdistinctapprox

Class KMVSketch

java.lang.Object

org.apache.sysds.runtime.matrix.data.sketch.countdistinctapprox.KMVSketch

All Implemented Interfaces:

MatrixSketch<Integer>

public class KMVSketch
extends Object

KMV synopsis(for k minimum values) Distinct-Value Estimation Kevin S. Beyer, Peter J. Haas, Berthold Reinwald, Yannis Sismanis, Rainer Gemulla: On synopses for distinct‐value estimation under multiset operations. SIGMOD 2007 TODO: Add multi-threaded version

Constructor Summary

Constructors

Constructor	Description
KMVSketch(Operator op)

Method Summary

Modifier and Type	Method	Description
CorrMatrixBlock	create(MatrixBlock blkIn)	Create a initial sketch of a given block.
MatrixBlock	getMatrixValue(CorrMatrixBlock arg0)	Obtain matrix distinct count value from estimation Used for estimating distinct in rows or columns.
Integer	getScalarValue(MatrixBlock in)	Get scalar distinct count from a input matrix block.
CorrMatrixBlock	intersection(CorrMatrixBlock arg0, CorrMatrixBlock arg1)	Intersect two sketches
CorrMatrixBlock	union(CorrMatrixBlock arg0, CorrMatrixBlock arg1)	Union two sketches together to from a combined sketch.
void	unionSketchByIndex(CorrMatrixBlock arg0, CorrMatrixBlock arg1, int idx, CorrMatrixBlock blkOut)

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

KMVSketch

public KMVSketch(Operator op)

Method Detail

getScalarValue

public Integer getScalarValue(MatrixBlock in)

Description copied from interface: MatrixSketch

Get scalar distinct count from a input matrix block.

Parameters:

in - A input block to estimate the number of distinct values in

Returns:

The distinct count estimate

getMatrixValue

public MatrixBlock getMatrixValue(CorrMatrixBlock arg0)

Description copied from interface: MatrixSketch

Obtain matrix distinct count value from estimation Used for estimating distinct in rows or columns.

Parameters:

arg0 - The sketch block to extract the count from

Returns:

The result matrix block

create

public CorrMatrixBlock create(MatrixBlock blkIn)

Description copied from interface: MatrixSketch

Create a initial sketch of a given block.

Parameters:

blkIn - A block to process

Returns:

A sketch

union

public CorrMatrixBlock union(CorrMatrixBlock arg0,
                             CorrMatrixBlock arg1)

Description copied from interface: MatrixSketch

Union two sketches together to from a combined sketch.

Parameters:

arg0 - Sketch one

arg1 - Sketch two

Returns:

The combined sketch

unionSketchByIndex

public void unionSketchByIndex(CorrMatrixBlock arg0,
                               CorrMatrixBlock arg1,
                               int idx,
                               CorrMatrixBlock blkOut)

intersection

public CorrMatrixBlock intersection(CorrMatrixBlock arg0,
                                    CorrMatrixBlock arg1)

Description copied from interface: MatrixSketch

Intersect two sketches

Parameters:

arg0 - Sketch one

arg1 - Sketch two

Returns:

The intersected sketch