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

Class LogarithmicBarrier

java.lang.Object

fr.cnes.sirius.patrius.math.optim.joptimizer.functions.LogarithmicBarrier

All Implemented Interfaces:

BarrierFunction, TwiceDifferentiableMultivariateRealFunction

public class LogarithmicBarrier
extends Object
implements BarrierFunction

Default barrier function for the barrier method algorithm.
If f_i(x) are the inequalities of the problem, theh we have:
Φ = - Sum_i[log(-f_i(x))]

Since:

4.6

Author:

alberto trivellato (alberto.trivellato@gmail.com)

See Also:

"S.Boyd and L.Vandenberghe, Convex Optimization, 11.2.1"

Constructor Summary

Constructors

Constructor and Description
LogarithmicBarrier(ConvexMultivariateRealFunction[] f, int dimension) Create the logarithmic barrier function.

Method Summary

Modifier and Type	Method and Description
double	calculatePhase1InitialFeasiblePoint(double[] originalNotFeasiblePoint, double tolerance) Calculates the initial value for the s parameter in Phase I.
BarrierFunction	createPhase1BarrierFunction() Create the barrier function for the Phase I.
int	getDim() Get dimension
double	getDualityGap(double t) Calculates the duality gap for a barrier method build with this barrier function
double[]	gradient(double[] value) Function gradient at point X
double[][]	hessian(double[] value) Function hessian at point X.
double	value(double[] val) Evaluation of the function at point X.

Methods inherited from class java.lang.Object

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

Constructor Detail

LogarithmicBarrier

public LogarithmicBarrier(ConvexMultivariateRealFunction[] f,
                          int dimension)

Create the logarithmic barrier function.

Parameters:

f - convex multivariate function

dimension - dimension

See Also:

"S.Boyd and L.Vandenberghe, Convex Optimization, 11.2.1"

Method Detail

value

public double value(double[] val)

Evaluation of the function at point X.

Specified by:

value in interface TwiceDifferentiableMultivariateRealFunction

Parameters:

val - point

Returns:

evaluation

gradient

public double[] gradient(double[] value)

Function gradient at point X

Specified by:

gradient in interface TwiceDifferentiableMultivariateRealFunction

Parameters:

value - point

Returns:

gradient

hessian

public double[][] hessian(double[] value)

Function hessian at point X.

Specified by:

hessian in interface TwiceDifferentiableMultivariateRealFunction

Parameters:

value - point

Returns:

hessian

getDim

public int getDim()

Get dimension

Specified by:

getDim in interface TwiceDifferentiableMultivariateRealFunction

Returns:

dimension

getDualityGap

public double getDualityGap(double t)

Calculates the duality gap for a barrier method build with this barrier function

Specified by:

getDualityGap in interface BarrierFunction

Parameters:

t - value

Returns:

duality gap

createPhase1BarrierFunction

public BarrierFunction createPhase1BarrierFunction()

Create the barrier function for the Phase I. It is a LogarithmicBarrier for the constraints:
fi(X)-s, i=1,...,n

Specified by:

createPhase1BarrierFunction in interface BarrierFunction

Returns:

phase 1 barrier function

See Also:

"S.Boyd and L.Vandenberghe, Convex Optimization, 11.4.1"

calculatePhase1InitialFeasiblePoint

public double calculatePhase1InitialFeasiblePoint(double[] originalNotFeasiblePoint,
                                                  double tolerance)

Calculates the initial value for the s parameter in Phase I. Return s = max(fi(x))

Specified by:

calculatePhase1InitialFeasiblePoint in interface BarrierFunction

Parameters:

originalNotFeasiblePoint - initial point (not-feasible)

tolerance - tolerance

Returns:

phase 1 initial feasible point

See Also:

"S.Boyd and L.Vandenberghe, Convex Optimization, 11.6.2"