algorithmlbg.java_2

包含了模式识别中常用的一些分类器设计算法

//----------------------------------------------------------------------
// AlgorithmLBG.java
//
// Created: 7/15/03
//
// Author: Phil Trasatti
// 
// Last Edited by: Daniel May
//
// Class: AlgorithmLBG
// 
// 
//---------------------------------------------------------------------

//----------------------
// import java packages
//----------------------
import java.util.*;
import java.awt.*;


public class AlgorithmLBG extends Algorithm
{

    //-----------------------------------------------------------------
    //
    // static data members
    //
    //-----------------------------------------------------------------

    //-----------------------------------------------------------------
    //
    // primitive data members
    //
    //-----------------------------------------------------------------

    int output_canvas_d[][];   
    int iterations;
    int currObject;
    int clusters;

    //-----------------------------------------------------------------
    //
    // instance data members
    //
    //-----------------------------------------------------------------
        
    String algo_id = "AlgorithmLBG"; 
    Vector support_vectors_d;
    Vector data_pool_d;
    Vector cbinary_d;
    DecisionRegion decision_regions_d;
    OutputPanel plain_data;

    //---------------------------------------------------------------
    //
    // class methods
    //
    //---------------------------------------------------------------


    //---------------------------------------------------------------
    // method: initialize
    //
    // arguments: none
    // return   : boolean
    //
    // description:  
    // overrides the initialize() method in the base class.  initializes
    // member data and prepares for execution of first step.  this method
    // "resets" the algorithm.
    //---------------------------------------------------------------
    public boolean initialize()
    {
	// Debug
	System.out.println(algo_id + ": initialize()");
	
	scale = output_panel_d.disp_area_d.getDisplayScale();
	data_pool_d = new Vector();
	decision_regions_d = new DecisionRegion();
	point_means_d = new Vector();
	support_vectors_d = new Vector();
	description_d = new Vector();
	step_count = 3;
	iterations = clusters;
	System.out.println("" + iterations);
	guesses_d = new Vector(2, 2);
	cbinary_d = new Vector();
	currObject = 0;

	// add the process description for the LBG algorithm
	if (description_d.size() == 0)
	    {
		String str = new String("   0. Initialize the original data.");
		description_d.addElement(str);
		
		str = new String("   1. Displaying the original data.");
		description_d.addElement(str);
		
		str = new String("   2. Computing the means for each class.");
		description_d.addElement(str);
		
		str = new String("   3. Stepping through iterations.");
		description_d.addElement(str);

		str = new String("Iteration");
		description_d.addElement(str);		
	    }
	
	// append message to process box
	pro_box_d.appendMessage("LBG Analysis:" + "\n");
	
	// set the data points for this algorithm
	set1_d = (Vector)data_points_d.dset1.clone();
	set2_d = (Vector)data_points_d.dset2.clone();
	set3_d = (Vector)data_points_d.dset3.clone();
	set4_d = (Vector)data_points_d.dset4.clone();
	
	// advance to step 1
	step_index_d = 0;
	
	// append message to process box
	pro_box_d.appendMessage((String)description_d.get(step_index_d));
		
	// exit gracefully
	return true;
	
    }

    //---------------------------------------------------------------
    // method: step1
    //
    // arguments: none
    // return   : boolean
    //
    // description:  
    // displays data sets from input box in output box.
    // 
    //---------------------------------------------------------------
    boolean step1()
    {
	// Debug
	System.out.println(algo_id + " : step1()");
	
	// set up progress bar
	pro_box_d.setProgressMin(0);
	pro_box_d.setProgressMax(1);
	pro_box_d.setProgressCurr(0);
	
	scaleToFitData();

	// Display original data
	output_panel_d.addOutput(set1_d, Classify.PTYPE_INPUT, 
				 data_points_d.color_dset1);
	output_panel_d.addOutput(set2_d, Classify.PTYPE_INPUT,
				 data_points_d.color_dset2);
	output_panel_d.addOutput(set3_d, Classify.PTYPE_INPUT,
				 data_points_d.color_dset3);
	output_panel_d.addOutput(set4_d, Classify.PTYPE_INPUT, 
				 data_points_d.color_dset4);
	
	plain_data = output_panel_d.copy();

	// step 1 completed
	pro_box_d.setProgressCurr(1);
	output_panel_d.repaint();
	
	return true;
    }

    //---------------------------------------------------------------
    // method: step2
    //
    // arguments: none
    // return   : boolean
    //
    // description:  
    // computes the means of each data set and displays the means graphically
    // and numerically
    // 
    //---------------------------------------------------------------
    boolean step2()
    {
	// Debug
	System.out.println(algo_id + " : step2()");
	
	// determine the within class scatter matrix
	generatePool();
	decision_regions_d.addRegion((Vector)data_pool_d.clone());

	MyPoint mean = MathUtil.computeClusterMean(data_pool_d);

	computeMeans();

	guesses_d.addElement(new MyPoint(mean.x, mean.y));

	decision_regions_d.setGuesses((Vector)guesses_d.clone());


	output_panel_d.addOutput(decision_regions_d.getGuesses(), 
				 Classify.PTYPE_OUTPUT_LARGE, 
				 Color.black);

	//cbinary_d.addElement(decision_regions_d);	

	pro_box_d.setProgressCurr(20);
	output_panel_d.repaint();
	
	return true;
    }
    
    //---------------------------------------------------------------
    // method: step3
    //
    // arguments: none
    // return   : boolean
    //
    // description:   
    // Computes the Decision Regions and the associated errors. 
    // 
    //---------------------------------------------------------------    
    boolean step3()
    {
	// Debug
	System.out.println(algo_id + " : step3()");
	
	if (currObject == 0)
	    pro_box_d.appendMessage((String)description_d.get(step_index_d));

	if (currObject < iterations)
	{
	    step_index_d--;

	    output_panel_d.clear();

	    output_panel_d = plain_data.copy();

	    output_panel_d.addOutput(set1_d, Classify.PTYPE_INPUT, 
	    		     data_points_d.color_dset1);
	    output_panel_d.addOutput(set2_d, Classify.PTYPE_INPUT,
	    		     data_points_d.color_dset2);
	    output_panel_d.addOutput(set3_d, Classify.PTYPE_INPUT,
	    		     data_points_d.color_dset3);
	    output_panel_d.addOutput(set4_d, Classify.PTYPE_INPUT, 
	                   data_points_d.color_dset4);
		    

	    output_panel_d.addOutput(decision_regions_d.getGuesses(), 
				     Classify.PTYPE_OUTPUT_LARGE, 
				     Color.black);


	    guesses_d = decision_regions_d.getGuesses();

	    System.out.println("Num Guesses" + guesses_d.size());

	    Vector guesses_d = decision_regions_d.;

	    Vector decisionRegions = decision_regions_d.getRegions();

	    //DecisionRegion region = new DecisionRegion(decision_regions_d);

	    decision_regions_d = new DecisionRegion();

	    // classify the data based on the guesses
	    //
	    classify(decision_regions_d);
	    
	    // compute the new means of the classified data
	    //
	    computeBinaryDeviates(decision_regions_d);
	    
	    // store the current iteration of the data set
	    //
	    //cbinary_d.addElement(decision_regions_d);
	    
	    output_panel_d.repaint();

	    currObject++;

	    pro_box_d.appendMessage("   Displaying Iteration: " + currObject  
				    + "\n");

	    //DecisionRegion region;

	    //region = (DecisionRegion)cbinary_d.elementAt(currObject);
	    //   System.out.println("Size of cbinary " + cbinary_d.size());

 	    // determine the total number of clusters
	    //
	    int numclusters = decision_regions_d.getNumRegions();
	    
	    // determine the classification error of each cluster
	    //
	    double error = 0.0;
	    double total = 0.0;   

	    System.out.println("numclusters: " + numclusters);
	    
	    for (int i=0; i<numclusters; i++) {
		
		// get the cluster associated with the region
		//
		Vector cluster = decision_regions_d.getRegion(i);
		
		// determine the mean of the current cluster
		//
		MyPoint mean = MathUtil.computeClusterMean(cluster);
		
		System.out.println(mean.toString());
		// display the mean for the current cluster
		//
		//double xval = setDecimal((mean.x) + Xmin, 2);
		//double yval = setDecimal(Ymax - (mean.y), 2);
		
		String message = new String("      Mean for cluster " +
					    i + ": " 
					    + mean.x + ", " + mean.y);
		pro_box_d.appendMessage(message + "\n");
		
		// determine the covariance matrix for the cluster
		//
		double x[] = new double[cluster.size()];
		double y[] = new double[cluster.size()];
		
		for (int j=0; j<cluster.size(); j++) {
		    MyPoint covarPoint =  (MyPoint)cluster.elementAt(j);
		    x[j] = (covarPoint.x);
		    y[j] = (covarPoint.y);
		}
		
		// declare the covariance object
		//
		Covariance cov = new Covariance();
		
		// declare the covariance matrix
		//
		Matrix covar = new Matrix();
		covar.row = covar.col = 2;
		covar.Elem = new double[2][2];
		
		// compute the covariance matrix of the first data set
		//
		covar.Elem = cov.computeCovariance(x, y);
		
		// display the covariance matrix for the cluster
		//
		double c11 = MathUtil.setDecimal(covar.Elem[0][0], 2);
		double c12 = MathUtil.setDecimal(covar.Elem[0][1], 2);
		double c21 = MathUtil.setDecimal(covar.Elem[1][0], 2);
		double c22 = MathUtil.setDecimal(covar.Elem[1][1], 2);
		
		message = new String("      Covariance matrix:\n" +
				     "         " + c11 
				     + "    " + c12 + "\n" +
				     "         " + c21 
				     + "    " + c22);
		
		pro_box_d.appendMessage(message + "\n");
		
		// determine the closest of the original data sets
		//
		int closest = getClosestSet(mean);
		
		// compute the classification error
		//
		total += (double)cluster.size();
		error += (double)displayClusterError(closest, 
						     cluster, i);
	    }
	    
	    double err = (error / total) * 100.0;
	    
	    // display the clasification error
	    //
	    String message =  new String("       Overall results:\n" +
					 "          Total number of samples: " + 
					 total + "\n" +
					 "          Misclassified samples: " + 
					 error + "\n" +
					 "          Classification error: " + 
					 MathUtil.setDecimal(err, 2) + "%");
	    pro_box_d.appendMessage(message + "\n");
	    
	    // update the output canvas
	    //
	    output_panel_d.repaint();
	}
	return true;
    }

    //---------------------------------------------------------------
    // method: run
    //
    // arguments: none
    // return   : none
    //
    // description: 
    // implementation of the run function from the Runnable interface.
    // determines what the current step is and calls the appropriate method.
    //--------------------------------------------------------------- 
    public void run()
    {
	// Debug
	System.out.println(algo_id + " : run()");
	
	if (step_index_d == 1)
	    {
		step1();
	    }
	
	else if (step_index_d == 2)
	    {
		step2(); 
	    }
	
	else if (step_index_d == 3)
	    {
		step3();
	    }
	// exit gracefully
	//
	return;
    }

    public void generatePool() {

	// determine the pool of points only once
	//
	if (data_pool_d.size() > 0) {