maxentclassifierfactory.java

Standord Classifier实现了一个基于Java的最大熵分类器。用于模式识别

package edu.stanford.nlp.classify;

import java.util.*;
import edu.stanford.nlp.util.*;
import edu.stanford.nlp.optimization.*;


/** @author Dan Klein */

public class MaxentClassifierFactory extends AbstractLinearClassifierFactory {

  double TOL = 1e-4;
  double sigma = 1.0;
  boolean verbose = false;

  public double[][] trainWeights(int[][] data, int[] labels) {
    Minimizer minimizer = new CGMinimizer(!verbose);
    DiffFunction objective = new LogConditionalObjectiveFunction(numFeatures(), numClasses(), data, labels, sigma);
    double[] weights = minimizer.minimize(objective, TOL, ((HasInitial)objective).initial());
    //System.err.println("Final derivative norm2: "+((AbstractCachingDiffFunction)objective).norm2(((AbstractCachingDiffFunction)objective).derivative));
    return ((LogConditionalObjectiveFunction)objective).to2D(weights);
  }

  /** Set the smoothing amount: 1.0 is the default. */
  public void setSigma(double sigma) {
    this.sigma = sigma;
  }

  public void setTol(double tol) {
    TOL = tol;
  }

  public void setVerbose(boolean verbose) {
    this.verbose = verbose;
  }

}