com.github.signaflo.math.optim

Class BFGS

java.lang.Object

com.github.signaflo.math.optim.BFGS

public final class BFGS
extends Object

An implementation of the Broyden-Fletcher-Goldfarb-Shanno (BFGS) algorithm for unconstrained nonlinear optimization. This class is immutable and thread-safe.

Constructor Summary

Constructors

Constructor and Description
BFGS(AbstractMultivariateFunction f, Vector startingPoint, double gradientTolerance, double functionChangeTolerance) Create a new BFGS object and run the algorithm with the supplied information.
BFGS(AbstractMultivariateFunction f, Vector startingPoint, double gradientNormTolerance, double relativeChangeTolerance, Matrix initialHessian) Create a new BFGS object and run the algorithm with the supplied information.

Method Summary

Modifier and Type	Method and Description
double	functionValue() Return the final value of the target function.
Matrix	inverseHessian() Return the final approximation to the inverse Hessian.
Vector	parameters() Return the final, optimized input parameters.

Methods inherited from class java.lang.Object

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

Constructor Detail

BFGS

public BFGS(AbstractMultivariateFunction f,
            Vector startingPoint,
            double gradientTolerance,
            double functionChangeTolerance)

Create a new BFGS object and run the algorithm with the supplied information.

Parameters:

f - the function to be minimized.

startingPoint - the initial guess of the minimum.

gradientTolerance - the tolerance for the norm of the gradient of the function.

functionChangeTolerance - the tolerance for the change in function value.

BFGS

public BFGS(AbstractMultivariateFunction f,
            Vector startingPoint,
            double gradientNormTolerance,
            double relativeChangeTolerance,
            Matrix initialHessian)

Create a new BFGS object and run the algorithm with the supplied information.

Parameters:

f - the function to be minimized.

startingPoint - the initial guess of the minimum.

gradientNormTolerance - the tolerance for the norm of the gradient of the function.

relativeChangeTolerance - the tolerance for the change in function value.

initialHessian - The initial guess for the inverse Hessian approximation.

Method Detail

functionValue

public double functionValue()

Return the final value of the target function.

Returns:

the final value of the target function.

parameters

public Vector parameters()

Return the final, optimized input parameters.

Returns:

the final, optimized input parameters.

inverseHessian

public Matrix inverseHessian()

Return the final approximation to the inverse Hessian.

Returns:

the final approximation to the inverse Hessian.