simplecriterion.java

一个很好的LIBSVM的JAVA源码。对于要研究和改进SVM算法的学者。可以参考。来自数据挖掘工具YALE工具包。

/*
 *  YALE - Yet Another Learning Environment
 *  Copyright (C) 2001-2004
 *      Simon Fischer, Ralf Klinkenberg, Ingo Mierswa, 
 *          Katharina Morik, Oliver Ritthoff
 *      Artificial Intelligence Unit
 *      Computer Science Department
 *      University of Dortmund
 *      44221 Dortmund,  Germany
 *  email: [e-mail]
 *  web:   http://yale.cs.uni-dortmund.de/
 *
 *  This program is free software; you can redistribute it and/or
 *  modify it under the terms of the GNU General Public License as 
 *  published by the Free Software Foundation; either version 2 of the
 *  License, or (at your option) any later version. 
 *
 *  This program is distributed in the hope that it will be useful, but
 *  WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
 *  General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with this program; if not, write to the Free Software
 *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
 *  USA.
 */
package edu.udo.cs.yale.operator.performance;

import edu.udo.cs.yale.tools.math.Averagable;
import edu.udo.cs.yale.example.Example;
import edu.udo.cs.yale.example.ExampleSet;
import edu.udo.cs.yale.operator.OperatorException;

/** Simple criteria are those which error can be counted for each example and can be
 *  averaged by the number of examples. 
 *  The fitness has a reciprocal value of the error.
 *
 *  @author Ingo, Simon
 *  @version $Id: SimpleCriterion.java,v 2.11 2004/08/27 11:57:43 ingomierswa Exp $
 */
public abstract class SimpleCriterion extends MeasuredPerformance {

    private double sum        = 0.0;
    private double squaresSum = 0.0;
    private int exampleCount  = 0;

    /** Invokes <code>countExample(double, double)</code> and uses 0 as predicted label if it not a number (NaN). */
    public void countExample(Example example) {
	double plabel = example.getPredictedLabel();
	if (Double.isNaN(plabel)) plabel = 0;
	double deviation = countExample(example.getLabel(), plabel);
	if (!Double.isNaN(deviation)) {
	    countExample(deviation);
	}
    }

    /** Subclasses must count the example and return the value to sum up. */
    protected abstract double countExample(double label, double predictedLabel);

    protected void countExample(double deviation) {
	if (!Double.isNaN(deviation)) {
	    sum        += deviation;
	    squaresSum += deviation*deviation;
	    exampleCount++;
	}
    }    

    public double getValue() {
	return sum / exampleCount;
    } 

    public double getVariance() {
	double mean        = getValue();
	double meanSquares = squaresSum / exampleCount;
	return meanSquares - mean*mean;
    }

    public void startCounting(ExampleSet eset) throws OperatorException {
	super.startCounting(eset);
	exampleCount = 0;
	sum = squaresSum = 0;
    }

    public double getFitness() {
	if (getValue() == 0.0) return Double.POSITIVE_INFINITY;
	return 1/getValue();
    }
    
    protected void cloneAveragable(Averagable newAvg) {
	//super.cloneAveragable(newAvg);
	SimpleCriterion sc = (SimpleCriterion)newAvg;
	this.sum          = sc.sum;
	this.squaresSum   = sc.squaresSum;
	this.exampleCount = sc.exampleCount;
    }

    public void buildAverage(Averagable performance) {
	super.buildAverage(performance);
	SimpleCriterion other = (SimpleCriterion)performance;
	this.sum          += other.sum;
	this.squaresSum   += other.squaresSum;
	this.exampleCount += other.exampleCount;
    }
}