gov.sandia.cognition.learning.algorithm.root.RootFinderRiddersMethod Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of cognitive-foundry Show documentation
Show all versions of cognitive-foundry Show documentation
A single jar with all the Cognitive Foundry components.
/*
* File: RootFinderRiddersMethod.java
* Authors: Kevin R. Dixon
* Company: Sandia National Laboratories
* Project: Cognitive Foundry
*
* Copyright Feb 6, 2009, Sandia Corporation.
* Under the terms of Contract DE-AC04-94AL85000, there is a non-exclusive
* license for use of this work by or on behalf of the U.S. Government.
* Export of this program may require a license from the United States
* Government. See CopyrightHistory.txt for complete details.
*
*/
package gov.sandia.cognition.learning.algorithm.root;
import gov.sandia.cognition.annotation.PublicationReference;
import gov.sandia.cognition.annotation.PublicationType;
import gov.sandia.cognition.learning.algorithm.minimization.line.InputOutputSlopeTriplet;
/**
* The root-finding algorithm due to Ridders. This algorithm is guaranteed
* to find a root if given a valid bracket. It tends to have very good
* performance on real-world cases. It works by performing stable logarithmic
* interpolation between function evaluations in a way that maintains a
* rigorous bracket on a root. It has good overall performance and will
* converge to a root.
* @author Kevin R. Dixon
* @since 3.0
*/
@PublicationReference(
author="Wikipedia",
title="Ridders' Method",
type=PublicationType.WebPage,
year=2009,
url="http://en.wikipedia.org/wiki/Ridders%27_method"
)
public class RootFinderRiddersMethod
extends AbstractBracketedRootFinder
{
/**
* Creates a new instance of RootFinderRiddersMethod
*/
public RootFinderRiddersMethod()
{
}
@Override
protected boolean step()
{
double x1 = this.getRootBracket().getLowerBound().getInput();
double f1 = this.getRootBracket().getLowerBound().getOutput();
double x2 = this.getRootBracket().getUpperBound().getInput();
double f2 = this.getRootBracket().getUpperBound().getOutput();
double xmid = 0.5 * (x1+x2);
double fmid = this.data.evaluate( xmid );
InputOutputSlopeTriplet midpoint =
new InputOutputSlopeTriplet( xmid, fmid, null );
double t1 = Math.sqrt( fmid*fmid - f1*f2 );
if( t1 == 0.0 )
{
return false;
}
double xnew = xmid + (xmid-x1)*Math.signum(f1-f2)*fmid / t1;
double fnew = this.data.evaluate( xnew );
InputOutputSlopeTriplet newpoint =
new InputOutputSlopeTriplet( xnew, fnew, null );
this.getRootBracket().setOtherPoint( newpoint );
if( fnew == 0.0 )
{
return false;
}
// If the midpoint and the new point have opposite signs, then we've
// found a much tighter bracket!!
if( fmid * fnew < 0.0 )
{
this.getRootBracket().setLowerBound( midpoint );
this.getRootBracket().setUpperBound( newpoint );
}
// If the new point has the same sign as the lower bound, then we
// can throw away the old lower bound
else if( f1*fnew > 0.0 )
{
this.getRootBracket().setLowerBound( newpoint );
}
// If the new point has the same sign as the upper bound, then
// we can throw away the old upper bound.
else if( f2*fnew > 0.0 )
{
this.getRootBracket().setUpperBound( newpoint );
}
else
{
throw new IllegalArgumentException( "You should never get here: "
+ "f1: " + f1 + ", f2: " + f2 + ", fnew: " + fnew + ", fmid: " + fmid );
}
double delta = this.getRootBracket().getLowerBound().getInput()
- this.getRootBracket().getUpperBound().getInput();
return Math.abs(delta) > this.getTolerance();
}
}