net.myrrix.common.math

Class MatrixUtils

java.lang.Object

net.myrrix.common.math.MatrixUtils

public final class MatrixUtils
extends Object

Contains utility methods for dealing with matrices, which are here represented as FastByIDMaps of FastByIDFloatMaps, or of float[].

Author:

Sean Owen

Method Summary

Methods

Modifier and Type	Method and Description
static void	addTo(long row, long column, float value, FastByIDMap<FastByIDFloatMap> RbyRow, FastByIDMap<FastByIDFloatMap> RbyColumn) Efficiently increments an entry in two parallel, sparse matrices.
static FastByIDMap<float[]>	getPseudoInverse(FastByIDMap<float[]> M) Pseudo-inverts a tall skinny matrix M.
static org.apache.commons.math3.linear.RealMatrix	getTransposeTimesSelfInverse(FastByIDMap<float[]> M)
static org.apache.commons.math3.linear.RealMatrix	invert(org.apache.commons.math3.linear.RealMatrix M)
static String	matrixToString(FastByIDMap<FastByIDFloatMap> M)
static FastByIDMap<float[]>	multiply(org.apache.commons.math3.linear.RealMatrix M, FastByIDMap<float[]> S)
static double[]	multiply(org.apache.commons.math3.linear.RealMatrix matrix, float[] V)
static org.apache.commons.math3.linear.RealMatrix	multiplyXYT(FastByIDMap<float[]> X, FastByIDMap<float[]> Y)
static void	remove(long row, long column, FastByIDMap<FastByIDFloatMap> RbyRow, FastByIDMap<FastByIDFloatMap> RbyColumn) Efficiently removes an entry in two parallel, sparse matrices.
static org.apache.commons.math3.linear.RealMatrix	transposeTimesSelf(FastByIDMap<float[]> M)

Methods inherited from class java.lang.Object

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

Method Detail

addTo

public static void addTo(long row,
         long column,
         float value,
         FastByIDMap<FastByIDFloatMap> RbyRow,
         FastByIDMap<FastByIDFloatMap> RbyColumn)

Efficiently increments an entry in two parallel, sparse matrices.

Parameters:

row - row to increment

column - column to increment

value - increment value

RbyRow - matrix R to update, keyed by row

RbyColumn - matrix R to update, keyed by column

remove

public static void remove(long row,
          long column,
          FastByIDMap<FastByIDFloatMap> RbyRow,
          FastByIDMap<FastByIDFloatMap> RbyColumn)

Efficiently removes an entry in two parallel, sparse matrices.

Parameters:

row - row to remove

column - column to remove

RbyRow - matrix R to update, keyed by row

RbyColumn - matrix R to update, keyed by column

getPseudoInverse

public static FastByIDMap<float[]> getPseudoInverse(FastByIDMap<float[]> M)

Pseudo-inverts a tall skinny matrix M. The result can be used as a left-inverse of M or right-inverse of MT:

((MT * M)^-1 * MT) * M = I

MT * (M * (MT * M)^-1) = I

Parameters:

M - tall skinny matrix

Returns:

a pseudo-inverse of M

getTransposeTimesSelfInverse

public static org.apache.commons.math3.linear.RealMatrix getTransposeTimesSelfInverse(FastByIDMap<float[]> M)

invert

public static org.apache.commons.math3.linear.RealMatrix invert(org.apache.commons.math3.linear.RealMatrix M)

multiply

public static FastByIDMap<float[]> multiply(org.apache.commons.math3.linear.RealMatrix M,
                            FastByIDMap<float[]> S)

Parameters:

M - small RealMatrix

S - wide, short matrix

Returns:

M * S as a newly allocated matrix

multiplyXYT

public static org.apache.commons.math3.linear.RealMatrix multiplyXYT(FastByIDMap<float[]> X,
                                                     FastByIDMap<float[]> Y)

multiply

public static double[] multiply(org.apache.commons.math3.linear.RealMatrix matrix,
                float[] V)

transposeTimesSelf

public static org.apache.commons.math3.linear.RealMatrix transposeTimesSelf(FastByIDMap<float[]> M)

Parameters:

M - tall, skinny matrix

Returns:

MT * M as a dense matrix

matrixToString

public static String matrixToString(FastByIDMap<FastByIDFloatMap> M)

Parameters:

M - matrix to print

Returns:

a print-friendly rendering of a sparse matrix. Not useful for wide matrices.