beam.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.util;

import java.util.*;

/**
 * Implements a finite beam, taking a comparator (default is
 * ScoredComparator.ASCENDING_COMPARATOR, the MAX object according to
 * the comparator is the one to be removed) and a beam size on
 * construction (default is 100).  Adding an object may cause the
 * worst-scored object to be removed from the beam (and that object
 * may well be the newly added object itself.
 *
 * @author <a href="mailto:[e-mail]">Dan Klein</a>
 * @version 1.0
 */
public class Beam extends AbstractSet {

    protected int maxBeamSize;
    protected Heap elements;

    public int capacity() { return maxBeamSize; }
    public int size() { return elements.size(); }

    public Iterator iterator() { return asSortedList().iterator(); }

  public List asSortedList() {
    LinkedList list = new LinkedList();
    for (Iterator i = elements.iterator(); i.hasNext();) {
      list.addFirst(i.next());
    }
    return list;
  }

    public boolean add(Object o) {
	boolean added = true;
	elements.add(o);
	while (size() > capacity()) {
	  Object dumped = elements.extractMin();
	  if (dumped.equals(o))
	    added = false;
	}
	return added;
    }

    public boolean remove(Object o) {
      //return elements.remove(o);
      throw new UnsupportedOperationException();
    }

    public Beam() {
	this(100);
    }
    
    public Beam(int maxBeamSize) {
	this(maxBeamSize, ScoredComparator.ASCENDING_COMPARATOR);
    }

    public Beam(int maxBeamSize, Comparator cmp) {
	elements = new ArrayHeap(cmp);
	this.maxBeamSize  = maxBeamSize;
    }

  public static void main(String[] args) {
    Beam b = new Beam(2, ScoredComparator.ASCENDING_COMPARATOR);
    b.add(new ScoredObject("1",1.0));
    b.add(new ScoredObject("2",2.0));
    b.add(new ScoredObject("3",3.0));
    b.add(new ScoredObject("0",0.0));
    for (Iterator bI = b.iterator(); bI.hasNext();) {
      ScoredObject sO = (ScoredObject)bI.next();
      System.out.println(sO);
    }
  }

}