Package org.ejml.dense.row.linsol.svd

Class SolvePseudoInverseSvd_FDRM

java.lang.Object

org.ejml.dense.row.linsol.svd.SolvePseudoInverseSvd_FDRM

All Implemented Interfaces:

org.ejml.interfaces.linsol.LinearSolver<org.ejml.data.FMatrixRMaj,org.ejml.data.FMatrixRMaj>, org.ejml.interfaces.linsol.LinearSolverDense<org.ejml.data.FMatrixRMaj>

@Generated("org.ejml.dense.row.linsol.svd.SolvePseudoInverseSvd_DDRM")
public class SolvePseudoInverseSvd_FDRM
extends Object
implements org.ejml.interfaces.linsol.LinearSolverDense<org.ejml.data.FMatrixRMaj>

The pseudo-inverse is typically used to solve over determined system for which there is no unique solution.
x=inv(A^TA)A^Tb
where A ∈ ℜ ^{m × n} and m ≥ n.

This class implements the Moore-Penrose pseudo-inverse using SVD and should never fail. Alternative implementations can use Cholesky decomposition, but those will fail if the A^TA matrix is singular. However the Cholesky implementation is much faster.

Constructor Summary

Constructors

Constructor	Description
SolvePseudoInverseSvd_FDRM()	Creates a solver targeted at matrices around 100x100
SolvePseudoInverseSvd_FDRM(int maxRows, int maxCols)	Creates a new solver targeted at the specified matrix size.

Method Summary

Modifier and Type	Method	Description
org.ejml.interfaces.decomposition.SingularValueDecomposition<org.ejml.data.FMatrixRMaj>	getDecomposer()
org.ejml.interfaces.decomposition.SingularValueDecomposition_F32<org.ejml.data.FMatrixRMaj>	getDecomposition()
void	invert(org.ejml.data.FMatrixRMaj A_inv)
boolean	modifiesA()
boolean	modifiesB()
double	quality()
boolean	setA(org.ejml.data.FMatrixRMaj A)
void	setThreshold(float threshold)	Specify the relative threshold used to select singular values.
void	solve(org.ejml.data.FMatrixRMaj b, org.ejml.data.FMatrixRMaj x)

Methods inherited from class java.lang.Object

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

Constructor Details

SolvePseudoInverseSvd_FDRM

public SolvePseudoInverseSvd_FDRM(int maxRows,
 int maxCols)

Creates a new solver targeted at the specified matrix size.

Parameters:

maxRows - The expected largest matrix it might have to process. Can be larger.

maxCols - The expected largest matrix it might have to process. Can be larger.

SolvePseudoInverseSvd_FDRM

public SolvePseudoInverseSvd_FDRM()

Creates a solver targeted at matrices around 100x100

Method Details

setA

public boolean setA(org.ejml.data.FMatrixRMaj A)

Specified by:

setA in interface org.ejml.interfaces.linsol.LinearSolver<org.ejml.data.FMatrixRMaj,org.ejml.data.FMatrixRMaj>

quality

public double quality()

Specified by:

quality in interface org.ejml.interfaces.linsol.LinearSolver<org.ejml.data.FMatrixRMaj,org.ejml.data.FMatrixRMaj>

solve

public void solve(org.ejml.data.FMatrixRMaj b,
 org.ejml.data.FMatrixRMaj x)

Specified by:

solve in interface org.ejml.interfaces.linsol.LinearSolver<org.ejml.data.FMatrixRMaj,org.ejml.data.FMatrixRMaj>

invert

public void invert(org.ejml.data.FMatrixRMaj A_inv)

Specified by:

invert in interface org.ejml.interfaces.linsol.LinearSolverDense<org.ejml.data.FMatrixRMaj>

modifiesA

public boolean modifiesA()

Specified by:

modifiesA in interface org.ejml.interfaces.linsol.LinearSolver<org.ejml.data.FMatrixRMaj,org.ejml.data.FMatrixRMaj>

modifiesB

public boolean modifiesB()

Specified by:

modifiesB in interface org.ejml.interfaces.linsol.LinearSolver<org.ejml.data.FMatrixRMaj,org.ejml.data.FMatrixRMaj>

getDecomposition

public org.ejml.interfaces.decomposition.SingularValueDecomposition_F32<org.ejml.data.FMatrixRMaj> getDecomposition()

Specified by:

getDecomposition in interface org.ejml.interfaces.linsol.LinearSolver<org.ejml.data.FMatrixRMaj,org.ejml.data.FMatrixRMaj>

setThreshold

public void setThreshold(float threshold)

Specify the relative threshold used to select singular values. By default it's UtilEjml.F_EPS.

Parameters:

threshold - The singular value threshold

getDecomposer

public org.ejml.interfaces.decomposition.SingularValueDecomposition<org.ejml.data.FMatrixRMaj> getDecomposer()