Package org.ejml.dense.row.mult

Class MatrixVectorMult_FDRM

java.lang.Object

org.ejml.dense.row.mult.MatrixVectorMult_FDRM

public class MatrixVectorMult_FDRM
extends java.lang.Object

This class contains various types of matrix vector multiplcation operations for FMatrixRMaj.

If a matrix has only one column or row then it is a vector. There are faster algorithms that can be used to multiply matrices by vectors. Strangely, even though the operations count smaller, the difference between this and a regular matrix multiply is insignificant for large matrices. The smaller matrices there is about a 40% speed improvement. In practice the speed improvement for smaller matrices is not noticeable unless 10s of millions of matrix multiplications are being performed.

Constructor Summary

Constructors

Constructor	Description
MatrixVectorMult_FDRM()

Method Summary

Modifier and Type	Method	Description
static float	innerProduct(float[] a, int offsetA, org.ejml.data.FMatrix1Row B, float[] c, int offsetC)	scalar = AT*B*C
static void	mult(org.ejml.data.FMatrix1Row A, org.ejml.data.FMatrixD1 B, org.ejml.data.FMatrixD1 C)	Performs a matrix vector multiply. c = A * b and c = A * bT ci = Sum{ j=1:n, aij * bj} where A is a matrix, b is a column or transposed row vector, and c is a column vector.
static void	multAdd(org.ejml.data.FMatrix1Row A, org.ejml.data.FMatrixD1 B, org.ejml.data.FMatrixD1 C)	Performs a matrix vector multiply. C = C + A * B or C = C + A * BT ci = Sum{ j=1:n, ci + aij * bj} where A is a matrix, B is a column or transposed row vector, and C is a column vector.
static void	multAddTransA_reorder(org.ejml.data.FMatrix1Row A, org.ejml.data.FMatrixD1 B, org.ejml.data.FMatrixD1 C)	An alternative implementation of multAddTransA_small(org.ejml.data.FMatrix1Row, org.ejml.data.FMatrixD1, org.ejml.data.FMatrixD1) that performs well on large matrices.
static void	multAddTransA_small(org.ejml.data.FMatrix1Row A, org.ejml.data.FMatrixD1 B, org.ejml.data.FMatrixD1 C)	Performs a matrix vector multiply. C = C + AT * B or C = CT + AT * BT ci = Sum{ j=1:n, ci + aji * bj} where A is a matrix, B is a column or transposed row vector, and C is a column vector.
static void	multTransA_reorder(org.ejml.data.FMatrix1Row A, org.ejml.data.FMatrixD1 B, org.ejml.data.FMatrixD1 C)	An alternative implementation of multTransA_small(org.ejml.data.FMatrix1Row, org.ejml.data.FMatrixD1, org.ejml.data.FMatrixD1) that performs well on large matrices.
static void	multTransA_small(org.ejml.data.FMatrix1Row A, org.ejml.data.FMatrixD1 B, org.ejml.data.FMatrixD1 C)	Performs a matrix vector multiply. C = AT * B where B is a column vector. or C = AT * BT where B is a row vector.

Methods inherited from class java.lang.Object

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

Constructor Detail

MatrixVectorMult_FDRM

public MatrixVectorMult_FDRM()

Method Detail

mult

public static void mult(org.ejml.data.FMatrix1Row A,
                        org.ejml.data.FMatrixD1 B,
                        org.ejml.data.FMatrixD1 C)

Performs a matrix vector multiply.

c = A * b
and
c = A * b^T

c_i = Sum{ j=1:n, a_ij * b_j}

where A is a matrix, b is a column or transposed row vector, and c is a column vector.

Parameters:

A - A matrix that is m by n. Not modified.

B - A vector that has length n. Not modified.

C - A column vector that has length m. Modified.

multAdd

public static void multAdd(org.ejml.data.FMatrix1Row A,
                           org.ejml.data.FMatrixD1 B,
                           org.ejml.data.FMatrixD1 C)

Performs a matrix vector multiply.

C = C + A * B
or
C = C + A * B^T

c_i = Sum{ j=1:n, c_i + a_ij * b_j}

where A is a matrix, B is a column or transposed row vector, and C is a column vector.

Parameters:

A - A matrix that is m by n. Not modified.

B - A vector that has length n. Not modified.

C - A column vector that has length m. Modified.

multTransA_small

public static void multTransA_small(org.ejml.data.FMatrix1Row A,
                                    org.ejml.data.FMatrixD1 B,
                                    org.ejml.data.FMatrixD1 C)

Performs a matrix vector multiply.

C = A^T * B
where B is a column vector.
or
C = A^T * B^T
where B is a row vector.

c_i = Sum{ j=1:n, a_ji * b_j}

where A is a matrix, B is a column or transposed row vector, and C is a column vector.

This implementation is optimal for small matrices. There is a huge performance hit when used on large matrices due to CPU cache issues.

Parameters:

A - A matrix that is m by n. Not modified.

B - A that has length m and is a column. Not modified.

C - A column vector that has length n. Modified.

multTransA_reorder

public static void multTransA_reorder(org.ejml.data.FMatrix1Row A,
                                      org.ejml.data.FMatrixD1 B,
                                      org.ejml.data.FMatrixD1 C)

An alternative implementation of multTransA_small(org.ejml.data.FMatrix1Row, org.ejml.data.FMatrixD1, org.ejml.data.FMatrixD1) that performs well on large matrices. There is a relative performance hit when used on small matrices.

Parameters:

A - A matrix that is m by n. Not modified.

B - A Vector that has length m. Not modified.

C - A column vector that has length n. Modified.

multAddTransA_small

public static void multAddTransA_small(org.ejml.data.FMatrix1Row A,
                                       org.ejml.data.FMatrixD1 B,
                                       org.ejml.data.FMatrixD1 C)

Performs a matrix vector multiply.

C = C + A^T * B
or
C = C^T + A^T * B^T

c_i = Sum{ j=1:n, c_i + a_ji * b_j}

where A is a matrix, B is a column or transposed row vector, and C is a column vector.

This implementation is optimal for small matrices. There is a huge performance hit when used on large matrices due to CPU cache issues.

Parameters:

A - A matrix that is m by n. Not modified.

B - A vector that has length m. Not modified.

C - A column vector that has length n. Modified.

multAddTransA_reorder

public static void multAddTransA_reorder(org.ejml.data.FMatrix1Row A,
                                         org.ejml.data.FMatrixD1 B,
                                         org.ejml.data.FMatrixD1 C)

An alternative implementation of multAddTransA_small(org.ejml.data.FMatrix1Row, org.ejml.data.FMatrixD1, org.ejml.data.FMatrixD1) that performs well on large matrices. There is a relative performance hit when used on small matrices.

Parameters:

A - A matrix that is m by n. Not modified.

B - A vector that has length m. Not modified.

C - A column vector that has length n. Modified.

innerProduct

public static float innerProduct(float[] a,
                                 int offsetA,
                                 org.ejml.data.FMatrix1Row B,
                                 float[] c,
                                 int offsetC)

scalar = A^T*B*C

Parameters:

a - (Input) vector

offsetA - Input) first index in vector a

B - (Input) Matrix

c - (Output) vector

offsetC - (Output) first index in vector c