matlabica.java

wekaUT是 university texas austin 开发的基于weka的半指导学习(semi supervised learning)的分类器

/*
 *    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.
 */

/*
 *    MatlabICA.java
 *    Copyright (C) 2002 Sugato Basu and Mikhail Bilenko 
 *
 */

package weka.attributeSelection;

import  java.io.*;
import  java.util.*;
import  weka.core.*;
import  weka.filters.unsupervised.attribute.ReplaceMissingValues;
import  weka.filters.unsupervised.attribute.Normalize;
import  weka.filters.unsupervised.attribute.NominalToBinary;
import  weka.filters.unsupervised.attribute.Remove;
import  weka.filters.Filter;

/**
 * Class for performing independent components analysis/transformation. <p>
 *
 * Valid options are:<p>
 * -D <br>
 * Don't normalize the input data. <p>
 *
 * -T <br>
 * Transform through the IC space and back to the original space. <p>
 * 
 * -N <br> num
 * Number of independant components
 *
 * -A <br> approach
 * ICA Approach
 *
 * -F <br> function
 * ICA function
 *
 * @author Sugato Basu
 * @author Mikhail Bilenko
 * @version $Revision: 1.2 $
 */
public class MatlabICA extends AttributeEvaluator 
  implements AttributeTransformer, OptionHandler {
  
  /** The data to transform analyse/transform */
  private Instances m_trainInstances;

  /** Keep a copy for the class attribute (if set) */
  private Instances m_trainCopy;

  /** The header for the transformed data format */
  private Instances m_transformedFormat;

  /** The header for data transformed back to the original space */
  private Instances m_originalSpaceFormat;

  /** Data has a class set */
  private boolean m_hasClass;

  /** Class index */
  private int m_classIndex;

  /** Number of attributes */
  private int m_numAttribs;

  /** Number of instances */
  private int m_numInstances;

  /** Name of the Matlab program file that computes ICA */
  protected String m_ICAMFile = new String("/var/local/MatlabICA.m");

  /** Will hold the mixing matrix */
  protected double [][] m_mixingMatrix;

  /** Will hold the inverse of the mixing matrix */
  protected double [][] m_inverseMixingMatrix;

  /** Will hold the independent components */
  protected double [][] m_independentComponents;

  /** A timestamp suffix for matching vectors with attributes */
  String m_timestamp = null;

  /** Name of the file where attribute names will be stored */
  String m_icaAttributeFilename = null;
  /** Name of the file where attribute names will be stored */
  String m_icaAttributeFilenameBase = new String("/var/local/ICAattributes");
  
  /** Name of the file where dataMatrix will be stored */
  public String m_dataFilename = new String("/var/local/ICAdataMatrix.txt");
  
  /** Name of the file where mixingMatrix will be stored */
  public String m_mixingMatrixFilename = null;
  public String m_mixingMatrixFilenameBase = new String("/var/local/ICAmixingMatrix");

  /** Name of the file where inverseMixingMatrix will be stored */
  public String m_inverseMixingMatrixFilename = new String("/var/local/ICAinverseMixingMatrix.txt");

  /** Name of the file where independentComponents will be stored */
  public String m_independentComponentsFilename = null;
  public String m_independentComponentsFilenameBase = new String("/var/local/ICAindependentComponents");
    
  /** Filters for original data */
  private ReplaceMissingValues m_replaceMissingFilter;
  private Normalize m_normalizeFilter;
  private Remove m_attributeFilter;
  
  /** The number of attributes in the ic transformed data */
  private int m_outputNumAtts = -1;
  
  /** normalize the input data? */
  private boolean m_normalize = true;

  /** transform the data through the ic space and back to the original
      space ? */
  private boolean m_transBackToOriginal = false;

  /** The attribute evaluator to use */
  private ASEvaluation m_eval = new weka.attributeSelection.ChiSquaredAttributeEval();

  /** load eigenvalues of covariance matrix from file? */
  protected boolean m_loadEigenValuesFromFile = false;

  /** set m_loadEigenValuesFromFile */
  public void setLoadEigenValuesFromFile(boolean choice) {
    m_loadEigenValuesFromFile = choice;
  }
  
  /** get m_loadEigenValuesFromFile */
  public boolean getLoadEigenValuesFromFile () {
    return m_loadEigenValuesFromFile;
  }

  /** load eigenvectors of covariance matrix from file? */
  protected boolean m_loadEigenVectorsFromFile = false;

  /** set m_loadEigenVectorsFromFile */
  public void setLoadEigenVectorsFromFile(boolean choice) {
    m_loadEigenVectorsFromFile = choice;
  }
  
  /** get m_loadEigenVectorsFromFile */
  public boolean getLoadEigenVectorsFromFile () {
    return m_loadEigenVectorsFromFile;
  }

  /** number of Independent Components */
  protected int m_NumIndependentComponents = 2;

  /** set number of Independent Components */
  public void setNumIndependentComponents(int n) {
    m_NumIndependentComponents = n;
    System.out.println("Number of ICA components: " + n);
  }

  /** get number of Independent Components */
  public int getNumIndependentComponents() {
    return m_NumIndependentComponents;
  }

  /* Define possible ICA approaches */
  public static final int APPROACH_SYMM = 0;
  public static final int APPROACH_DEFL = 1;
  public static final Tag[] TAGS_APPROACH = {
    new Tag(APPROACH_SYMM, "symm"),
    new Tag(APPROACH_DEFL, "defl")
  };
  protected int m_ICAapproach = APPROACH_SYMM;

  /** get ICA approach */
  public SelectedTag getICAapproach ()
  {
    return new SelectedTag(m_ICAapproach, TAGS_APPROACH);
  }

  /** set ICA approach */
  public void setICAapproach (SelectedTag approach)
  {
    if (approach.getTags() == TAGS_APPROACH) {
      System.out.println("Approach: " + approach.getSelectedTag().getReadable());
      m_ICAapproach = approach.getSelectedTag().getID();
    }
  }

  /* Define possible ICA functions */
  public static final int FUNCTION_TANH = 0;
  public static final int FUNCTION_GAUSS = 1;
  public static final int FUNCTION_POW3 = 2;
  public static final int FUNCTION_SKEW = 3;
  public static final Tag[] TAGS_FUNCTION = {
    new Tag(FUNCTION_TANH, "tanh"),
    new Tag(FUNCTION_GAUSS, "gauss"),
    new Tag(FUNCTION_POW3, "pow3"),
    new Tag(FUNCTION_SKEW, "skew")
  };
  protected int m_ICAfunction = FUNCTION_TANH;

  /** get ICA function */
  public SelectedTag getICAfunction ()
  {
    return new SelectedTag(m_ICAfunction, TAGS_FUNCTION);
  }

  /** set ICA function */
  public void setICAfunction (SelectedTag function)
  {
    if (function.getTags() == TAGS_FUNCTION) {
      System.out.println("Function: " + function.getSelectedTag().getReadable());
      m_ICAfunction = function.getSelectedTag().getID();
    }
  }

  /**
   * Returns a string describing this attribute transformer
   * @return a description of the evaluator suitable for
   * displaying in the explorer/experimenter gui
   */
  public String globalInfo() {
    return "Performs a independent components analysis and transformation of "
      +"the data. Use in conjunction with a Ranker search. Dimensionality "
      +"reduction is accomplished by choosing enough eigenvectors to "
      +"account for some percentage of the variance in the original data---"
      +"default 0.95 (95%). Attribute noise can be filtered by transforming "
      +"to the IC space, eliminating some of the worst eigenvectors, and "
      +"then transforming back to the original space.";
  }

  /**
   * Returns an enumeration describing the available options. <p>
   *
   * @return an enumeration of all the available options.
   **/
  public Enumeration listOptions () {
    Vector newVector = new Vector(3);
    newVector.addElement(new Option("\tDon't normalize input data." 
				    , "D", 0, "-D"));
    
    newVector.addElement(new Option("\tTransform through the IC space and "
				    +"\n\tback to the original space."
				    , "O", 0, "-O"));

    newVector.addElement(new Option("\tNumber of independant components." 
				    , "N", 1, "-N"));

    newVector.addElement(new Option("\tICA approach." 
				    , "A", 1, "-A"));

    newVector.addElement(new Option("\tICA function." 
				    , "F", 1, "-F"));

    return  newVector.elements();
  }

  /**
   * Parses a given list of options.
   *
   * Valid options are:<p>
   * -D <br>
   * Don't normalize the input data. <p>
   *
   * -T <br>
   * Transform through the IC space and back to the original space. <p>
   *
   * -N <br> num
   * Number of independant components
   *
   * -A <br> approach
   * ICA Approach
   *
   * -F <br> function
   * ICA function
   *
   * @param options the list of options as an array of strings
   * @exception Exception if an option is not supported
   */
  public void setOptions (String[] options)
    throws Exception
  {
    resetOptions();
    String optionString;

    setNormalize(!Utils.getFlag('D', options));
    setTransformBackToOriginal(Utils.getFlag('O', options));

    optionString = Utils.getOption('F', options);
    if (optionString.length() != 0) {
      setICAfunction(new SelectedTag(Integer.parseInt(optionString), TAGS_FUNCTION));
    }
    optionString = Utils.getOption('A', options);
    if (optionString.length() != 0) {
      setICAapproach(new SelectedTag(Integer.parseInt(optionString), TAGS_APPROACH));
    }
    optionString = Utils.getOption('N', options);
    if (optionString.length() != 0) {
      setNumIndependentComponents(Integer.parseInt(optionString));
    }
  }

  /**
   * Reset to defaults
   */
  private void resetOptions() {
    m_normalize = false;
    m_transBackToOriginal = false;
    m_ICAfunction = 0;
    m_ICAapproach = 0;
    m_eval = new weka.attributeSelection.ChiSquaredAttributeEval();
    m_NumIndependentComponents = 2;
  }

  /**
   * Sets the attribute evaluator
   *
   * @param evaluator the evaluator with all options set.
   */
  public void setEvaluator(ASEvaluation evaluator) {
    m_eval = evaluator;
  }

  /**
   * Gets the attribute evaluator used
   *
   * @return the attribute evaluator
   */