org.apache.commons.math3.optim.nonlinear.scalar.gradient
Class NonLinearConjugateGradientOptimizer.IdentityPreconditioner

java.lang.Object
  org.apache.commons.math3.optim.nonlinear.scalar.gradient.NonLinearConjugateGradientOptimizer.IdentityPreconditioner

All Implemented Interfaces:

Preconditioner

Enclosing class:

NonLinearConjugateGradientOptimizer

public static class NonLinearConjugateGradientOptimizer.IdentityPreconditioner
extends Object
implements Preconditioner

Default identity preconditioner.

Constructor Summary

NonLinearConjugateGradientOptimizer.IdentityPreconditioner()

Method Summary

double[]

precondition(double[] variables, double[] r) Precondition a search direction.

Methods inherited from class java.lang.Object

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

Constructor Detail

NonLinearConjugateGradientOptimizer.IdentityPreconditioner

public NonLinearConjugateGradientOptimizer.IdentityPreconditioner()

Method Detail

precondition

public double[] precondition(double[] variables,
                             double[] r)

Precondition a search direction.

The returned preconditioned search direction must be computed fast or the algorithm performances will drop drastically. A classical approach is to compute only the diagonal elements of the hessian and to divide the raw search direction by these elements if they are all positive. If at least one of them is negative, it is safer to return a clone of the raw search direction as if the hessian was the identity matrix. The rationale for this simplified choice is that a negative diagonal element means the current point is far from the optimum and preconditioning will not be efficient anyway in this case.

Specified by:

precondition in interface Preconditioner

Parameters:

variables - current point at which the search direction was computed

r - raw search direction (i.e. opposite of the gradient)

Returns:

approximation of H^-1r where H is the objective function hessian