kullbackleiblerdistancecalculator.java

java实现的隐马尔科夫模型

/* jahmm package - v0.3.1 */

/*
 *  Copyright (c) 2004, Jean-Marc Francois.
 *
 *  This file is part of Jahmm.
 *  Jahmm is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation; either version 2 of the License, or
 *  (at your option) any later version.
 *
 *  Jahmm is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with Jahmm; if not, write to the Free Software
 *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 */

package be.ac.ulg.montefiore.run.jahmm.toolbox;

import java.util.*;
import be.ac.ulg.montefiore.run.jahmm.*;


/**
 * Computes the distance between HMMs.
 * <p>
 * The distance metric is similar to the Kullback-Leibler distance defined
 * on distributions.  More information can be found in
 * <i>A Probabilistic Distance Measure For HMMs</i> by
 * <i>Rabiner</i> and <i>Juang</i> (AT&T Technical Journal, vol. 64, Feb. 1985,
 * pages 391-408).
 * <p>
 * This distance measure is not symetric: <code>distance(hmm1, hmm2)</code>
 * is not necessary equal to <code>distance(hmm2, hmm1)</code>.  To get a
 * symetric distance definition, compute
 * <code>(distance(hmm1, hmm2) + distance(hmm2, hmm1)) / 2</code>.
 */
public class KullbackLeiblerDistanceCalculator {
    
    private double distance;
    private final int generatedSequencesLength = 10000;
    
    
    /**
     * Computes the Kullback-Leibler distance between two HMMs.
     *
     * @param hmm1 The first HMM against which the distance is computed.
     *             The distance is mesured with regard to this HMM (this must
     *             be defined since the Kullback-Leibler distance is not
     *             symetric).
     * @param hmm2 The second HMM against which the distance is computed.
     */
    public KullbackLeiblerDistanceCalculator(Hmm hmm1, Hmm hmm2) {
	distance = KLCompute(hmm1, hmm2);
    }
    
    
    private double KLCompute(Hmm hmm1, Hmm hmm2) {
	Vector oseq = new MarkovGenerator(hmm1).
	    observationSequence(generatedSequencesLength);
	
	return 
	    (new ForwardBackwardScaledCalculator(oseq, hmm1).lnProbability() -
	     new ForwardBackwardScaledCalculator(oseq, hmm2).lnProbability()) /
	    generatedSequencesLength;
    }
    
    
    /**
     * Returns the distance between <code>hmm1</code> and <code>hmm2</code> with
     * regard to <code>hmm1</code>, where <code>hmm1</code> and
     * <code>hmm2</code> are the HMM given to 
     * {@link #KullbackLeiblerDistanceCalculator(Hmm, Hmm) contructor}.
     *
     * @return The distance between the HMMs.
     */
    public double distance() {
	 return distance;
    }
}