loglossboost.java

Boosting算法软件包

package jboost.booster;

import jboost.controller.Configuration;

/**
 * AdaBoost-like implementation for the Log-Loss. The only differnece is weight
 * update.
 * 
 * @author Yoram Singer
 * @version $Header:
 *          /proj/gene/cvs-repository/jboost/src/jboost/booster/LogLossBoost.java,v
 *          1.2 2003/10/01 18:36:15 freund Exp $
 */
public class LogLossBoost extends AdaBoost {


  /**
   * Default constructors call AdaBoost constructors
   * Uses a value of 0.0 for the smoothing term 
   */
  public LogLossBoost() {
    super(0.0);
  }

  /**
   * Constructor that initializes the smoothing term
   * @param smooth the value to use for the smoothing term
   */
  LogLossBoost(double smooth) {
    super(smooth);
    init(new Configuration());
  }

  /**
   * AdaBoost's finalizeData is overridden since the initial weight initialized
   * differently (q_{0,i} = 1/2)
   */
  public void finalizeData() {
    finalizeData(0.5);
  }

  /**
   * Return the theoretical bound on the training error.
   */
  public double getBound() {
    final double oneOverLog2= 1.44269504088896340737;
    double Z= 0.;

    for (int i= 0; i < m_margins.length; i++)
      Z += Math.log(1 + Math.exp(-m_margins[i]));

    return oneOverLog2 * Z / m_margins.length;
  }


    public String getParamString() {
        String ret = String.format("None (LogLossBoost)");
        return ret;
    }

  
  /**
   * The LogLoss example weight is set to 1/(1 + e^(margin))
   */
  public double calculateWeight(double margin) {
  	return 1/(1 + Math.exp(margin));
  }

  /** 
   * Update the examples m_margins and m_weights using the
   * logistic loss computation
   * @param predictions the value of the prediction for each example
   * @param exampleIndex  the indexes of each example to update
   */
  public void update(Prediction[] predictions, int[][] exampleIndex) {
    // exchange pointers of new and old m_weights
    for (int i= 0; i < m_weights.length; i++) {
      m_oldWeights[i]= m_weights[i];
    }
   
    // update m_weights and m_margins
    for (int i= 0; i < exampleIndex.length; i++) {
      double p= predictions[i].getClassScores()[1];
      double[] value= new double[] { -p, p };
      int[] indexes= exampleIndex[i];
      for (int j= 0; j < indexes.length; j++) {
        int example= indexes[j];
        m_margins[example] += value[m_labels[example]];
        m_weights[example]= calculateWeight(m_margins[example]);
      }
    }
  }
}