• Home
• nz.ac.waikato.cms.weka
• weka-dev
• 3.9.5
• source code
• SimulatedAnnealing.java

×

Project download

Many resources are needed to download a project. Please understand that we have to compensate our server costs. Thank you in advance.
Project price only 1 $

You can buy this project and download/modify it how often you want.

weka.classifiers.bayes.net.search.global.SimulatedAnnealing Maven / Gradle / Ivy

/*
 *   This program 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 3 of the License, or
 *   (at your option) any later version.
 *
 *   This program 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 this program.  If not, see .
 */

/*
 * SimulatedAnnealing.java
 * Copyright (C) 2004-2012 University of Waikato, Hamilton, New Zealand
 * 
 */

package weka.classifiers.bayes.net.search.global;

import java.util.Collections;
import java.util.Enumeration;
import java.util.Random;
import java.util.Vector;

import weka.classifiers.bayes.BayesNet;
import weka.core.Instances;
import weka.core.Option;
import weka.core.RevisionUtils;
import weka.core.TechnicalInformation;
import weka.core.TechnicalInformation.Field;
import weka.core.TechnicalInformation.Type;
import weka.core.TechnicalInformationHandler;
import weka.core.Utils;

/**
 *  This Bayes Network learning algorithm uses the
 * general purpose search method of simulated annealing to find a well scoring
 * network structure.

 * 

 * For more information see:

 * 

 * R.R. Bouckaert (1995). Bayesian Belief Networks: from Construction to
 * Inference. Utrecht, Netherlands.
 * 

 * 
 * 
 *  BibTeX:
 * 
 * 
 * @phdthesis{Bouckaert1995,
 *    address = {Utrecht, Netherlands},
 *    author = {R.R. Bouckaert},
 *    institution = {University of Utrecht},
 *    title = {Bayesian Belief Networks: from Construction to Inference},
 *    year = {1995}
 * }
 * 
 * 

 * 
 * 
 *  Valid options are:
 * 

 * 
 * 
 * -A <float>
 *  Start temperature
 * 
 * 
 * 
 * -U <integer>
 *  Number of runs
 * 
 * 
 * 
 * -D <float>
 *  Delta temperature
 * 
 * 
 * 
 * -R <seed>
 *  Random number seed
 * 
 * 
 * 
 * -mbc
 *  Applies a Markov Blanket correction to the network structure, 
 *  after a network structure is learned. This ensures that all 
 *  nodes in the network are part of the Markov blanket of the 
 *  classifier node.
 * 
 * 
 * 
 * -S [LOO-CV|k-Fold-CV|Cumulative-CV]
 *  Score type (LOO-CV,k-Fold-CV,Cumulative-CV)
 * 
 * 
 * 
 * -Q
 *  Use probabilistic or 0/1 scoring.
 *  (default probabilistic scoring)
 * 
 * 
 * 
 * 
 * @author Remco Bouckaert ([email protected])
 * @version $Revision: 11267 $
 */
public class SimulatedAnnealing extends GlobalScoreSearchAlgorithm implements
  TechnicalInformationHandler {

  /** for serialization */
  static final long serialVersionUID = -5482721887881010916L;

  /** start temperature **/
  double m_fTStart = 10;

  /** change in temperature at every run **/
  double m_fDelta = 0.999;

  /** number of runs **/
  int m_nRuns = 10000;

  /** use the arc reversal operator **/
  boolean m_bUseArcReversal = false;

  /** random number seed **/
  int m_nSeed = 1;

  /** random number generator **/
  Random m_random;

  /**
   * Returns an instance of a TechnicalInformation object, containing detailed
   * information about the technical background of this class, e.g., paper
   * reference or book this class is based on.
   * 
   * @return the technical information about this class
   */
  @Override
  public TechnicalInformation getTechnicalInformation() {
    TechnicalInformation result;

    result = new TechnicalInformation(Type.PHDTHESIS);
    result.setValue(Field.AUTHOR, "R.R. Bouckaert");
    result.setValue(Field.YEAR, "1995");
    result.setValue(Field.TITLE,
      "Bayesian Belief Networks: from Construction to Inference");
    result.setValue(Field.INSTITUTION, "University of Utrecht");
    result.setValue(Field.ADDRESS, "Utrecht, Netherlands");

    return result;
  }

  /**
   * 
   * @param bayesNet the bayes net to use
   * @param instances the data to use
   * @throws Exception if something goes wrong
   */
  @Override
  public void search(BayesNet bayesNet, Instances instances) throws Exception {
    m_random = new Random(m_nSeed);

    // determine base scores
    double fCurrentScore = calcScore(bayesNet);

    // keep track of best scoring network
    double fBestScore = fCurrentScore;
    BayesNet bestBayesNet = new BayesNet();
    bestBayesNet.m_Instances = instances;
    bestBayesNet.initStructure();
    copyParentSets(bestBayesNet, bayesNet);

    double fTemp = m_fTStart;
    for (int iRun = 0; iRun < m_nRuns; iRun++) {
      boolean bRunSucces = false;
      double fDeltaScore = 0.0;
      while (!bRunSucces) {
        // pick two nodes at random
        int iTailNode = m_random.nextInt(instances.numAttributes());
        int iHeadNode = m_random.nextInt(instances.numAttributes());
        while (iTailNode == iHeadNode) {
          iHeadNode = m_random.nextInt(instances.numAttributes());
        }
        if (isArc(bayesNet, iHeadNode, iTailNode)) {
          bRunSucces = true;
          // either try a delete
          bayesNet.getParentSet(iHeadNode).deleteParent(iTailNode, instances);
          double fScore = calcScore(bayesNet);
          fDeltaScore = fScore - fCurrentScore;
          // System.out.println("Try delete " + iTailNode + "->" + iHeadNode +
          // " dScore = " + fDeltaScore);
          if (fTemp
            * Math
              .log((Math.abs(m_random.nextInt()) % 10000) / 10000.0 + 1e-100) < fDeltaScore) {
            // System.out.println("success!!!");
            fCurrentScore = fScore;
          } else {
            // roll back
            bayesNet.getParentSet(iHeadNode).addParent(iTailNode, instances);
          }
        } else {
          // try to add an arc
          if (addArcMakesSense(bayesNet, instances, iHeadNode, iTailNode)) {
            bRunSucces = true;
            double fScore = calcScoreWithExtraParent(iHeadNode, iTailNode);
            fDeltaScore = fScore - fCurrentScore;
            // System.out.println("Try add " + iTailNode + "->" + iHeadNode +
            // " dScore = " + fDeltaScore);
            if (fTemp
              * Math
                .log((Math.abs(m_random.nextInt()) % 10000) / 10000.0 + 1e-100) < fDeltaScore) {
              // System.out.println("success!!!");
              bayesNet.getParentSet(iHeadNode).addParent(iTailNode, instances);
              fCurrentScore = fScore;
            }
          }
        }
      }
      if (fCurrentScore > fBestScore) {
        copyParentSets(bestBayesNet, bayesNet);
      }
      fTemp = fTemp * m_fDelta;
    }

    copyParentSets(bayesNet, bestBayesNet);
  } // buildStructure

  /**
   * CopyParentSets copies parent sets of source to dest BayesNet
   * 
   * @param dest destination network
   * @param source source network
   */
  void copyParentSets(BayesNet dest, BayesNet source) {
    int nNodes = source.getNrOfNodes();
    // clear parent set first
    for (int iNode = 0; iNode < nNodes; iNode++) {
      dest.getParentSet(iNode).copy(source.getParentSet(iNode));
    }
  } // CopyParentSets

  /**
   * @return double
   */
  public double getDelta() {
    return m_fDelta;
  }

  /**
   * @return double
   */
  public double getTStart() {
    return m_fTStart;
  }

  /**
   * @return int
   */
  public int getRuns() {
    return m_nRuns;
  }

  /**
   * Sets the m_fDelta.
   * 
   * @param fDelta The m_fDelta to set
   */
  public void setDelta(double fDelta) {
    m_fDelta = fDelta;
  }

  /**
   * Sets the m_fTStart.
   * 
   * @param fTStart The m_fTStart to set
   */
  public void setTStart(double fTStart) {
    m_fTStart = fTStart;
  }

  /**
   * Sets the m_nRuns.
   * 
   * @param nRuns The m_nRuns to set
   */
  public void setRuns(int nRuns) {
    m_nRuns = nRuns;
  }

  /**
   * @return random number seed
   */
  public int getSeed() {
    return m_nSeed;
  } // getSeed

  /**
   * Sets the random number seed
   * 
   * @param nSeed The number of the seed to set
   */
  public void setSeed(int nSeed) {
    m_nSeed = nSeed;
  } // setSeed

  /**
   * Returns an enumeration describing the available options.
   * 
   * @return an enumeration of all the available options.
   */
  @Override
  public Enumeration listOptions() {
    Vector newVector = new Vector(4);

    newVector
      .addElement(new Option("\tStart temperature", "A", 1, "-A "));
    newVector
      .addElement(new Option("\tNumber of runs", "U", 1, "-U "));
    newVector
      .addElement(new Option("\tDelta temperature", "D", 1, "-D "));
    newVector
      .addElement(new Option("\tRandom number seed", "R", 1, "-R "));

    newVector.addAll(Collections.list(super.listOptions()));

    return newVector.elements();
  }

  /**
   * Parses a given list of options.
   * 

   * 
   *  Valid options are:
   * 

   * 
   * 
   * -A <float>
   *  Start temperature
   * 
   * 
   * 
   * -U <integer>
   *  Number of runs
   * 
   * 
   * 
   * -D <float>
   *  Delta temperature
   * 
   * 
   * 
   * -R <seed>
   *  Random number seed
   * 
   * 
   * 
   * -mbc
   *  Applies a Markov Blanket correction to the network structure, 
   *  after a network structure is learned. This ensures that all 
   *  nodes in the network are part of the Markov blanket of the 
   *  classifier node.
   * 
   * 
   * 
   * -S [LOO-CV|k-Fold-CV|Cumulative-CV]
   *  Score type (LOO-CV,k-Fold-CV,Cumulative-CV)
   * 
   * 
   * 
   * -Q
   *  Use probabilistic or 0/1 scoring.
   *  (default probabilistic scoring)
   * 
   * 
   * 
   * 
   * @param options the list of options as an array of strings
   * @throws Exception if an option is not supported
   */
  @Override
  public void setOptions(String[] options) throws Exception {
    String sTStart = Utils.getOption('A', options);
    if (sTStart.length() != 0) {
      setTStart(Double.parseDouble(sTStart));
    }
    String sRuns = Utils.getOption('U', options);
    if (sRuns.length() != 0) {
      setRuns(Integer.parseInt(sRuns));
    }
    String sDelta = Utils.getOption('D', options);
    if (sDelta.length() != 0) {
      setDelta(Double.parseDouble(sDelta));
    }
    String sSeed = Utils.getOption('R', options);
    if (sSeed.length() != 0) {
      setSeed(Integer.parseInt(sSeed));
    }
    super.setOptions(options);
  }

  /**
   * Gets the current settings of the search algorithm.
   * 
   * @return an array of strings suitable for passing to setOptions
   */
  @Override
  public String[] getOptions() {

    Vector options = new Vector();

    options.add("-A");
    options.add("" + getTStart());

    options.add("-U");
    options.add("" + getRuns());

    options.add("-D");
    options.add("" + getDelta());

    options.add("-R");
    options.add("" + getSeed());

    Collections.addAll(options, super.getOptions());

    return options.toArray(new String[0]);
  }

  /**
   * This will return a string describing the classifier.
   * 
   * @return The string.
   */
  @Override
  public String globalInfo() {
    return "This Bayes Network learning algorithm uses the general purpose search method "
      + "of simulated annealing to find a well scoring network structure.\n\n"
      + "For more information see:\n\n" + getTechnicalInformation().toString();
  } // globalInfo

  /**
   * @return a string to describe the TStart option.
   */
  public String TStartTipText() {
    return "Sets the start temperature of the simulated annealing search. "
      + "The start temperature determines the probability that a step in the 'wrong' direction in the "
      + "search space is accepted. The higher the temperature, the higher the probability of acceptance.";
  } // TStartTipText

  /**
   * @return a string to describe the Runs option.
   */
  public String runsTipText() {
    return "Sets the number of iterations to be performed by the simulated annealing search.";
  } // runsTipText

  /**
   * @return a string to describe the Delta option.
   */
  public String deltaTipText() {
    return "Sets the factor with which the temperature (and thus the acceptance probability of "
      + "steps in the wrong direction in the search space) is decreased in each iteration.";
  } // deltaTipText

  /**
   * @return a string to describe the Seed option.
   */
  public String seedTipText() {
    return "Initialization value for random number generator."
      + " Setting the seed allows replicability of experiments.";
  } // seedTipText

  /**
   * Returns the revision string.
   * 
   * @return the revision
   */
  @Override
  public String getRevision() {
    return RevisionUtils.extract("$Revision: 11267 $");
  }
} // SimulatedAnnealing