randomrbf.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

   * @return the default number of classes
   */
  protected int defaultNumClasses() {
    return 2;
  }

  /**
   * Sets the number of classes the dataset should have.
   * @param numClasses the new number of classes
   */
  public void setNumClasses(int numClasses) { 
    m_NumClasses = numClasses; 
  }

  /**
   * Gets the number of classes the dataset should have.
   * @return the number of classes the dataset should have
   */
  public int getNumClasses() { 
    return m_NumClasses; 
  }
  
  /**
   * Returns the tip text for this property
   * 
   * @return tip text for this property suitable for
   *         displaying in the explorer/experimenter gui
   */
  public String numClassesTipText() {
    return "The number of classes to generate.";
  }

  /**
   * returns the default number of centroids
   * 
   * @return the default number of centroids
   */
  protected int defaultNumCentroids() {
    return 50;
  }
  
  /**
   * Gets the number of centroids.
   *
   * @return the number of centroids.
   */
  public int getNumCentroids() { 
    return m_NumCentroids; 
  }
  
  /**
   * Sets the number of centroids to use.
   *
   * @param value the number of centroids to use.
   */
  public void setNumCentroids(int value) { 
    if (value > 0)
      m_NumCentroids = value; 
    else
      System.out.println("At least 1 centroid is necessary (provided: " 
          + value + ")!");
  }  
  
  /**
   * Returns the tip text for this property
   * 
   * @return tip text for this property suitable for
   *         displaying in the explorer/experimenter gui
   */
  public String numCentroidsTipText() {
    return "The number of centroids to use.";
  }

  /**
   * Return if single mode is set for the given data generator
   * mode depends on option setting and or generator type.
   * 
   * @return single mode flag
   * @throws Exception if mode is not set yet
   */
  public boolean getSingleModeFlag() throws Exception {
    return true;
  }

  /**
   * returns a random index based on the given proportions
   *
   * @param proportionArray     the proportions
   * @param random              the random number generator to use
   * @return the random index
   */
  protected int chooseRandomIndexBasedOnProportions(
      double[] proportionArray, Random random) {

    double      probSum;
    double      val;
    int         index;
    double      sum;

    probSum = Utils.sum(proportionArray);
    val     = random.nextDouble() * probSum;
    index   = 0;
    sum     = 0.0;
    
    while ((sum <= val) && (index < proportionArray.length))
      sum += proportionArray[index++];
    
    return index - 1;
  }

  /**
   * Initializes the format for the dataset produced. 
   * Must be called before the generateExample or generateExamples
   * methods are used.
   * Re-initializes the random number generator with the given seed.
   *
   * @return the format for the dataset 
   * @throws Exception if the generating of the format failed
   * @see  #getSeed()
   */
  public Instances defineDataFormat() throws Exception {
    int             i;
    int             j;
    FastVector      atts;
    FastVector      clsValues;
    Random          rand;

    m_Random = new Random(getSeed());
    rand     = getRandom();

    // number of examples is the same as given per option
    setNumExamplesAct(getNumExamples());

    // initialize centroids
    m_centroids       = new double[getNumCentroids()][getNumAttributes()];
    m_centroidClasses = new int[getNumCentroids()];
    m_centroidWeights = new double[getNumCentroids()];
    m_centroidStdDevs = new double[getNumCentroids()];

    for (i = 0; i < getNumCentroids(); i++) {
      for (j = 0; j < getNumAttributes(); j++)
        m_centroids[i][j] = rand.nextDouble();
      m_centroidClasses[i] = rand.nextInt(getNumClasses());
      m_centroidWeights[i] = rand.nextDouble();
      m_centroidStdDevs[i] = rand.nextDouble();
    }

    // initialize dataset format
    atts = new FastVector();
    for (i = 0; i < getNumAttributes(); i++)
      atts.addElement(new Attribute("a" + i));

    clsValues = new FastVector();
    for (i = 0; i < getNumClasses(); i++)
      clsValues.addElement("c" + i);
    atts.addElement(new Attribute("class", clsValues));
    
    m_DatasetFormat = new Instances(getRelationNameToUse(), atts, 0);
    
    return m_DatasetFormat;
  }

  /**
   * Generates one example of the dataset. 
   *
   * @return the generated example
   * @throws Exception if the format of the dataset is not yet defined
   * @throws Exception if the generator only works with generateExamples
   * which means in non single mode
   */
  public Instance generateExample() throws Exception {
    Instance    result;
    int         centroid;
    double[]    atts;
    double      magnitude;
    double      desiredMag;
    double      scale;
    int         i;
    double      label;
    Random      rand;

    result = null;
    rand   = getRandom();

    if (m_DatasetFormat == null)
      throw new Exception("Dataset format not defined.");

    // generate class label based on class probs
    centroid = chooseRandomIndexBasedOnProportions(m_centroidWeights, rand);
    label    = m_centroidClasses[centroid];

    // generate attributes
    atts = new double[getNumAttributes() + 1];
    for (i = 0; i < getNumAttributes(); i++)
      atts[i] = (rand.nextDouble() * 2.0) - 1.0;
    atts[atts.length - 1] = label;
    
    magnitude = 0.0;
    for (i = 0; i < getNumAttributes(); i++)
      magnitude += atts[i] * atts[i];
    
    magnitude  = Math.sqrt(magnitude);
    desiredMag = rand.nextGaussian() * m_centroidStdDevs[centroid];
    scale      = desiredMag / magnitude;
    for (i = 0; i < getNumAttributes(); i++) {
      atts[i] *= scale;
      atts[i] += m_centroids[centroid][i];
      result   = new Instance(1.0, atts);
    }

    // dataset reference
    result.setDataset(m_DatasetFormat);
    
    return result;
  }

  /**
   * Generates all examples of the dataset. Re-initializes the random number
   * generator with the given seed, before generating instances.
   *
   * @return the generated dataset
   * @throws Exception if the format of the dataset is not yet defined
   * @throws Exception if the generator only works with generateExample,
   * which means in single mode
   * @see   #getSeed()
   */
  public Instances generateExamples() throws Exception {
    Instances       result;
    int             i;

    result   = new Instances(m_DatasetFormat, 0);
    m_Random = new Random(getSeed());

    for (i = 0; i < getNumExamplesAct(); i++)
      result.add(generateExample());
    
    return result;
  }

  /**
   * Generates a comment string that documentates the data generator.
   * By default this string is added at the beginning of the produced output
   * as ARFF file type, next after the options.
   * 
   * @return string contains info about the generated rules
   */
  public String generateStart () {
    StringBuffer        result;
    int                 i;

    result = new StringBuffer();

    result.append("%\n");
    result.append("% centroids:\n");
    for (i = 0; i < getNumCentroids(); i++)
      result.append(
          "% " + i + ".: " + Utils.arrayToString(m_centroids[i]) + "\n");
    result.append("%\n");
    result.append(
        "% centroidClasses: " + Utils.arrayToString(m_centroidClasses) + "\n");
    result.append("%\n");
    result.append(
        "% centroidWeights: " + Utils.arrayToString(m_centroidWeights) + "\n");
    result.append("%\n");
    result.append(
        "% centroidStdDevs: " + Utils.arrayToString(m_centroidStdDevs) + "\n");
    result.append("%\n");
    
    return result.toString();
  }

  /**
   * Generates a comment string that documentats the data generator.
   * By default this string is added at the end of theproduces output
   * as ARFF file type.
   * 
   * @return string contains info about the generated rules
   * @throws Exception if the generating of the documentaion fails
   */
  public String generateFinished() throws Exception {
    return "";
  }

  /**
   * Main method for executing this class.
   *
   * @param args should contain arguments for the data producer: 
   */
  public static void main(String[] args) {
    runDataGenerator(new RandomRBF(), args);
  }
}