fr.cnes.sirius.patrius.math.optim.joptimizer.solvers

Class AbstractKKTSolver

java.lang.Object

fr.cnes.sirius.patrius.math.optim.joptimizer.solvers.AbstractKKTSolver

Direct Known Subclasses:

AugmentedKKTSolver, BasicKKTSolver, UpperDiagonalHKKTSolver

public abstract class AbstractKKTSolver
extends Object

Solves the KKT system: H.v + [A]T.w = -g,
A.v = -h,
(H is square and symmetric)

Since:

4.6

Author:

alberto trivellato (alberto.trivellato@gmail.com)

See Also:

"S.Boyd and L.Vandenberghe, Convex Optimization, p. 542"

Field Summary

Fields

Modifier and Type	Field and Description
protected boolean	checkKKTSolutionAcc Check if the KTT accuracy is preserved
static double	DEFAULT_SCALAR Default scalar
protected double	defaultScalar Default scalar
protected RealVector	g Vector g
protected RealVector	h Vector h
protected RealMatrix	matA Matrix A
protected RealMatrix	matAT Transpose A matrix
protected RealMatrix	matH Matrix H
protected double	toleranceKKT KTT tolerance

Constructor Summary

Constructors

Constructor and Description
AbstractKKTSolver()

Method Summary

Modifier and Type	Method and Description
protected boolean	checkKKTSolutionAccuracy(RealVector v, RealVector w) Check the solution of the system KKT.x = b against the scaled residual beta < gamma, where gamma is a parameter chosen by the user and beta is the scaled residual, beta = ||KKT.x-b||_oo/( ||KKT||_oo .
void	setAMatrix(RealMatrix aMatrix) Set the A matrix
void	setCheckKKTSolutionAccuracy(boolean b) Set if the KTT solution accuracy is preserved or not
void	setGVector(RealVector gVector) Set the g vector
void	setHMatrix(RealMatrix hMatrix) Set the H matrix
void	setHVector(RealVector hVector) Set the h vector
void	setToleranceKKT(double tolerance) Acceptable tolerance for system resolution.
abstract RealVector[]	solve() Returns two vectors v and w solutions of the KKT system.
protected RealVector[]	solveAugmentedKKT() Solve the augmented KTT system

Methods inherited from class java.lang.Object

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

Field Detail

DEFAULT_SCALAR

public static final double DEFAULT_SCALAR

Default scalar

See Also:

Constant Field Values

matH

protected RealMatrix matH

Matrix H

matA

protected RealMatrix matA

Matrix A

matAT

protected RealMatrix matAT

Transpose A matrix

g

protected RealVector g

Vector g

h

protected RealVector h

Vector h

toleranceKKT

protected double toleranceKKT

KTT tolerance

checkKKTSolutionAcc

protected boolean checkKKTSolutionAcc

Check if the KTT accuracy is preserved

defaultScalar

protected double defaultScalar

Default scalar

Constructor Detail

AbstractKKTSolver

public AbstractKKTSolver()

Method Detail

solve

public abstract RealVector[] solve()
                            throws PatriusException

Returns two vectors v and w solutions of the KKT system.

Returns:

v and w

Throws:

PatriusException - if an error occurs

setHMatrix

public void setHMatrix(RealMatrix hMatrix)

Set the H matrix

Parameters:

hMatrix - matrix H

setAMatrix

public void setAMatrix(RealMatrix aMatrix)

Set the A matrix

Parameters:

aMatrix - matrix A

setGVector

public void setGVector(RealVector gVector)

Set the g vector

Parameters:

gVector - vector G

setHVector

public void setHVector(RealVector hVector)

Set the h vector

Parameters:

hVector - vector H

setToleranceKKT

public void setToleranceKKT(double tolerance)

Acceptable tolerance for system resolution.

Parameters:

tolerance - KTT tolerance

setCheckKKTSolutionAccuracy

public void setCheckKKTSolutionAccuracy(boolean b)

Set if the KTT solution accuracy is preserved or not

Parameters:

b - true or false

solveAugmentedKKT

protected RealVector[] solveAugmentedKKT()
                                  throws PatriusException

Solve the augmented KTT system

Returns:

vector with the solution

Throws:

PatriusException - if an error occurs

checkKKTSolutionAccuracy

protected boolean checkKKTSolutionAccuracy(RealVector v,
                                           RealVector w)

Check the solution of the system KKT.x = b against the scaled residual beta < gamma, where gamma is a parameter chosen by the user and beta is the scaled residual, beta = ||KKT.x-b||_oo/( ||KKT||_oo . ||x||_oo + ||b||_oo ), with ||x||_oo = max(||x[i]||)

Parameters:

v - vector V

w - vector W

Returns:

true/false