miemdd.java

代码是一个分类器的实现,其中使用了部分weka的源代码。可以将项目导入eclipse运行

/*
 *    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 2 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, write to the Free Software
 *    Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

/*
 * MIEMDD.java
 * Copyright (C) 2005 University of Waikato, Hamilton, New Zealand
 *
 */

package weka.classifiers.mi;

import weka.classifiers.RandomizableClassifier;
import weka.core.Capabilities;
import weka.core.FastVector;
import weka.core.Instance;
import weka.core.Instances;
import weka.core.MultiInstanceCapabilitiesHandler;
import weka.core.Optimization;
import weka.core.Option;
import weka.core.OptionHandler;
import weka.core.SelectedTag;
import weka.core.Tag;
import weka.core.TechnicalInformation;
import weka.core.TechnicalInformationHandler;
import weka.core.Utils;
import weka.core.Capabilities.Capability;
import weka.core.TechnicalInformation.Field;
import weka.core.TechnicalInformation.Type;
import weka.filters.Filter;
import weka.filters.unsupervised.attribute.Normalize;
import weka.filters.unsupervised.attribute.ReplaceMissingValues;
import weka.filters.unsupervised.attribute.Standardize;

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

/**
 <!-- globalinfo-start -->
 * EMDD model builds heavily upon Dietterich's Diverse Density (DD) algorithm.<br/>
 * It is a general framework for MI learning of converting the MI problem to a single-instance setting using EM. In this implementation, we use most-likely cause DD model and only use 3 random selected postive bags as initial starting points of EM.<br/>
 * <br/>
 * For more information see:<br/>
 * <br/>
 * Qi Zhang, Sally A. Goldman: EM-DD: An Improved Multiple-Instance Learning Technique. In: Advances in Neural Information Processing Systems 14, 1073-108, 2001.
 * <p/>
 <!-- globalinfo-end -->
 * 
 <!-- technical-bibtex-start -->
 * BibTeX:
 * <pre>
 * &#64;inproceedings{Zhang2001,
 *    author = {Qi Zhang and Sally A. Goldman},
 *    booktitle = {Advances in Neural Information Processing Systems 14},
 *    pages = {1073-108},
 *    publisher = {MIT Press},
 *    title = {EM-DD: An Improved Multiple-Instance Learning Technique},
 *    year = {2001}
 * }
 * </pre>
 * <p/>
 <!-- technical-bibtex-end -->
 *
 <!-- options-start -->
 * Valid options are: <p/>
 * 
 * <pre> -N &lt;num&gt;
 *  Whether to 0=normalize/1=standardize/2=neither.
 *  (default 1=standardize)</pre>
 * 
 * <pre> -S &lt;num&gt;
 *  Random number seed.
 *  (default 1)</pre>
 * 
 * <pre> -D
 *  If set, classifier is run in debug mode and
 *  may output additional info to the console</pre>
 * 
 <!-- options-end -->
 *     
 * @author Eibe Frank ([e-mail])
 * @author Lin Dong ([e-mail])
 * @version $Revision: 1.5 $ 
 */
public class MIEMDD 
  extends RandomizableClassifier 
  implements OptionHandler, MultiInstanceCapabilitiesHandler,
             TechnicalInformationHandler {

  /** for serialization */
  static final long serialVersionUID = 3899547154866223734L;
  
  /** The index of the class attribute */
  protected int m_ClassIndex;

  protected double[] m_Par;

  /** The number of the class labels */
  protected int m_NumClasses;

  /** Class labels for each bag */
  protected int[] m_Classes;

  /** MI data */
  protected double[][][] m_Data;

  /** All attribute names */
  protected Instances m_Attributes;

  /** MI data */	
  protected double[][] m_emData;

  /** The filter used to standardize/normalize all values. */
  protected Filter m_Filter = null;

  /** Whether to normalize/standardize/neither, default:standardize */
  protected int m_filterType = FILTER_STANDARDIZE;

  /** Normalize training data */
  public static final int FILTER_NORMALIZE = 0;
  /** Standardize training data */
  public static final int FILTER_STANDARDIZE = 1;
  /** No normalization/standardization */
  public static final int FILTER_NONE = 2;
  /** The filter to apply to the training data */
  public static final Tag[] TAGS_FILTER = {
    new Tag(FILTER_NORMALIZE, "Normalize training data"),
    new Tag(FILTER_STANDARDIZE, "Standardize training data"),
    new Tag(FILTER_NONE, "No normalization/standardization"),
  };

  /** The filter used to get rid of missing values. */
  protected ReplaceMissingValues m_Missing = new ReplaceMissingValues();

  /**
   * Returns a string describing this filter
   *
   * @return a description of the filter suitable for
   * displaying in the explorer/experimenter gui
   */
  public String globalInfo() {
    return 
        "EMDD model builds heavily upon Dietterich's Diverse Density (DD) "
      + "algorithm.\nIt is a general framework for MI learning of converting "
      + "the MI problem to a single-instance setting using EM. In this "
      + "implementation, we use most-likely cause DD model and only use 3 "
      + "random selected postive bags as initial starting points of EM.\n\n"
      + "For more information see:\n\n"
      + getTechnicalInformation().toString();
  }

  /**
   * 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
   */
  public TechnicalInformation getTechnicalInformation() {
    TechnicalInformation 	result;
    
    result = new TechnicalInformation(Type.INPROCEEDINGS);
    result.setValue(Field.AUTHOR, "Qi Zhang and Sally A. Goldman");
    result.setValue(Field.TITLE, "EM-DD: An Improved Multiple-Instance Learning Technique");
    result.setValue(Field.BOOKTITLE, "Advances in Neural Information Processing Systems 14");
    result.setValue(Field.YEAR, "2001");
    result.setValue(Field.PAGES, "1073-108");
    result.setValue(Field.PUBLISHER, "MIT Press");
    
    return result;
  }

  /**
   * Returns an enumeration describing the available options
   *
   * @return an enumeration of all the available options
   */
  public Enumeration listOptions() {
    Vector result = new Vector();
    
    result.addElement(new Option(
          "\tWhether to 0=normalize/1=standardize/2=neither.\n" 
          + "\t(default 1=standardize)",
          "N", 1, "-N <num>"));

    Enumeration enm = super.listOptions();
    while (enm.hasMoreElements())
      result.addElement(enm.nextElement());

    return result.elements();
  }

  /**
   * Parses a given list of options. <p/>
   * 
   <!-- options-start -->
   * Valid options are: <p/>
   * 
   * <pre> -N &lt;num&gt;
   *  Whether to 0=normalize/1=standardize/2=neither.
   *  (default 1=standardize)</pre>
   * 
   * <pre> -S &lt;num&gt;
   *  Random number seed.
   *  (default 1)</pre>
   * 
   * <pre> -D
   *  If set, classifier is run in debug mode and
   *  may output additional info to the console</pre>
   * 
   <!-- options-end -->
   *
   * @param options the list of options as an array of strings
   * @throws Exception if an option is not supported
   */
  public void setOptions(String[] options) throws Exception {
    String 	tmpStr;
    
    tmpStr = Utils.getOption('N', options);
    if (tmpStr.length() != 0) {
      setFilterType(new SelectedTag(Integer.parseInt(tmpStr), TAGS_FILTER));
    } else {
      setFilterType(new SelectedTag(FILTER_STANDARDIZE, TAGS_FILTER));
    }     

    super.setOptions(options);
  }


  /**
   * Gets the current settings of the classifier.
   *
   * @return an array of strings suitable for passing to setOptions
   */
  public String[] getOptions() {
    Vector	result;
    String[]	options;
    int		i;
    
    result  = new Vector();
    options = super.getOptions();
    for (i = 0; i < options.length; i++)
      result.add(options[i]);
    
    result.add("-N");
    result.add("" + m_filterType);

    return (String[]) result.toArray(new String[result.size()]);
  }

  /**
   * Returns the tip text for this property
   *
   * @return tip text for this property suitable for
   * displaying in the explorer/experimenter gui
   */
  public String filterTypeTipText() {
    return "The filter type for transforming the training data.";
  }

  /**
   * Gets how the training data will be transformed. Will be one of
   * FILTER_NORMALIZE, FILTER_STANDARDIZE, FILTER_NONE.
   *
   * @return the filtering mode
   */
  public SelectedTag getFilterType() {
    return new SelectedTag(m_filterType, TAGS_FILTER);
  }

  /**
   * Sets how the training data will be transformed. Should be one of
   * FILTER_NORMALIZE, FILTER_STANDARDIZE, FILTER_NONE.
   *
   * @param newType the new filtering mode
   */
  public void setFilterType(SelectedTag newType) {

    if (newType.getTags() == TAGS_FILTER) {
      m_filterType = newType.getSelectedTag().getID();
    }
  }

  private class OptEng 
    extends Optimization {
    /**
     * Evaluate objective function
     * @param x the current values of variables
     * @return the value of the objective function
     */
    protected double objectiveFunction(double[] x){
      double nll = 0; // -LogLikelihood
      for (int i=0; i<m_Classes.length; i++){ // ith bag
        double ins=0.0;
        for (int k=0; k<m_emData[i].length; k++)  //attribute index
          ins += (m_emData[i][k]-x[k*2])*(m_emData[i][k]-x[k*2])*
            x[k*2+1]*x[k*2+1];
        ins = Math.exp(-ins); // Pr. of being positive

        if (m_Classes[i]==1){
          if (ins <= m_Zero) ins = m_Zero;
          nll -= Math.log(ins); //bag level -LogLikelihood
        }
        else{
          ins = 1.0 - ins;  //Pr. of being negative
          if(ins<=m_Zero) ins=m_Zero;
          nll -= Math.log(ins);
        }
      }
      return nll;
    }

    /**
     * Evaluate Jacobian vector
     * @param x the current values of variables
     * @return the gradient vector
     */
    protected double[] evaluateGradient(double[] x){
      double[] grad = new double[x.length];
      for (int i=0; i<m_Classes.length; i++){ // ith bag
        double[] numrt = new double[x.length];
        double exp=0.0;
        for (int k=0; k<m_emData[i].length; k++) //attr index
          exp += (m_emData[i][k]-x[k*2])*(m_emData[i][k]-x[k*2])
            *x[k*2+1]*x[k*2+1];
        exp = Math.exp(-exp);  //Pr. of being positive

        //Instance-wise update
        for (int p=0; p<m_emData[i].length; p++){  // pth variable
          numrt[2*p] = 2.0*(x[2*p]-m_emData[i][p])*x[p*2+1]*x[p*2+1];
          numrt[2*p+1] = 2.0*(x[2*p]-m_emData[i][p])*(x[2*p]-m_emData[i][p])
            *x[p*2+1];
        }

        //Bag-wise update
        for (int q=0; q<m_emData[i].length; q++){
          if (m_Classes[i] == 1) {//derivation of (-LogLikeliHood) for positive bags
            grad[2*q] += numrt[2*q];
            grad[2*q+1] += numrt[2*q+1];
          }
          else{ //derivation of (-LogLikeliHood) for negative bags
            grad[2*q] -= numrt[2*q]*exp/(1.0-exp);
            grad[2*q+1] -= numrt[2*q+1]*exp/(1.0-exp);
          }
        }
      } // one bag

      return grad;
    }
  }

  /**
   * Returns default capabilities of the classifier.
   *
   * @return      the capabilities of this classifier
   */
  public Capabilities getCapabilities() {
    Capabilities result = super.getCapabilities();

    // attributes