evolutionaryweighting.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.features.weighting;

import edu.udo.cs.yale.example.ExampleSet;
import edu.udo.cs.yale.example.Attribute;
import edu.udo.cs.yale.example.AttributeWeightedExampleSet;
import edu.udo.cs.yale.operator.IOObject;
import edu.udo.cs.yale.operator.features.*;
import edu.udo.cs.yale.operator.features.ga.*;
import edu.udo.cs.yale.operator.performance.PerformanceVector;
import edu.udo.cs.yale.operator.parameter.*;
import edu.udo.cs.yale.operator.OperatorException;
import edu.udo.cs.yale.tools.RandomGenerator;

import java.util.List;
import java.util.LinkedList;
import java.util.Random;

/** This operator performs the weighting of features with a evolutionary approach.
 *   
 *  @version $Id: EvolutionaryWeighting.java,v 1.12 2004/09/14 08:39:06 ingomierswa Exp $
 */
public class EvolutionaryWeighting extends FeatureOperator {

    /** The selection modes which are usable with this genetic algorithm. */
    private static final String[] SELECTION_MODES = {
	"keep_best_n", "roulette_wheel", "keep_best"
    };

    /** Indicates that the n best individuals should be used for the next generation. */
    private static final int KEEP_BEST_N    = 0;
    /** Indicates that a roulette wheel selection should be used for the next generation. */
    private static final int ROULETTE_WHEEL = 1;
    /** Indicates that only the best individual should be used for the next generation. */
    private static final int KEEP_BEST      = 2;

    /** The number of individuals in each population. */
    private int numberOfIndividuals;
    /** The maximum generation. */
    private int maxGeneration;
    /** The maximum number of generations without improval. */
    private int maxWithoutImproval;

    /** The list with pre-evaluation population operators. */
    private List preOps  = new LinkedList();
    /** The list with post-evaluation population operators. */
    private List postOps = new LinkedList();


    public IOObject[] apply() throws OperatorException {
	this.numberOfIndividuals = getParameterAsInt("population_size");
	this.maxGeneration       = getParameterAsInt("maximum_number_of_generations");
	this.maxWithoutImproval  = getParameterAsInt("generations_without_improval");

	preOps = new LinkedList();
	// selection
	int selectionMode = getParameterAsInt("selection_mode");
	switch (selectionMode) {
	    case KEEP_BEST_N:
		preOps.add(new KeepBest(numberOfIndividuals));
		break;
	    case ROULETTE_WHEEL:
		preOps.add(new RouletteWheel(numberOfIndividuals,
					     getParameterAsBoolean("keep_best_individual")));
		break;
	    case KEEP_BEST:
		preOps.add(new BestSelection());
		break;
	}

	// crossover & mutation
	preOps.add(new WeightingCrossover(getParameterAsInt("crossover_type"), 
					  getParameterAsDouble("p_crossover")));
	WeightingMutation weighting = new WeightingMutation(getParameterAsDouble("init_variance"), getParameterAsBoolean("bounded_mutation"));
	preOps.add(weighting);


	postOps = new LinkedList();
	// mutation adaption
	if (getParameterAsBoolean("1_5_rule")) 
	    postOps.add(new VarianceAdaption(weighting));

	return super.apply();
    }

    public boolean solutionGoodEnough(Population population) {
	return ((population.getGeneration() >= maxGeneration) || 
		(population.getGenerationsWithoutImproval() >= maxWithoutImproval));
    }

    public List getPreEvaluationPopulationOperators() {
	return preOps;
    }

    public List getPostEvaluationPopulationOperators() {
	return postOps;
    }

    public Population createInitialPopulation(ExampleSet exampleSet) {
	Population initPop = new Population();
	for (int i = 0; i < numberOfIndividuals; i++) {
	    AttributeWeightedExampleSet nes = new AttributeWeightedExampleSet((ExampleSet)exampleSet.clone());
	    for (int j = 0; j < nes.getNumberOfAttributes(); j++) {
		nes.setWeightForBlock(j, RandomGenerator.getGlobalRandomGenerator().nextDouble());
		//nes.setWeightForBlock(j, 1.0d);
	    }
	    initPop.add(nes);
	}
	return initPop;
    }

    public List getParameterTypes() {
	List types = super.getParameterTypes();
	ParameterType type = new ParameterTypeInt("population_size", "Number of individuals per generation.", 1, Integer.MAX_VALUE, 5);
	type.setExpert(false);
	types.add(type);
	type = new ParameterTypeInt("maximum_number_of_generations", 
				    "The maximum number of generations.", 
				    1, Integer.MAX_VALUE, 10);
	type.setExpert(false);
	types.add(type);
	type = new ParameterTypeInt("generations_without_improval", 
				    "The maximum number of generations without improval.", 
				    1, Integer.MAX_VALUE, 5);
	type.setExpert(false);
	types.add(type);
	types.add(new ParameterTypeCategory("selection_mode", "The selection mode for this genetic algorithm.", 
					    SELECTION_MODES, KEEP_BEST_N));
	types.add(new ParameterTypeBoolean("keep_best_individual", 
					   "If set to true, the best individual of each generations is guaranteed to be selected for the next generation (elitist selection, only used for roulette wheel).", false));
	types.add(new ParameterTypeBoolean("1_5_rule", 
					   "If set to true, the 1/5 rule for variance adaption is used.", true));
	types.add(new ParameterTypeBoolean("bounded_mutation", 
					   "If set to true, the weights are bounded between 0 and 1.", false));
	types.add(new ParameterTypeDouble("init_variance", "The init variance for each mutation.", 0.0d, 
					  Double.POSITIVE_INFINITY, 1.0d));
	type = new ParameterTypeDouble("p_crossover", 
				       "Probability for an individual to be selected for crossover.", 0, 1, 0.3);
	type.setExpert(false);
	types.add(type);
	types.add(new ParameterTypeCategory("crossover_type", 
					    "Type of the crossover.", 
					    WeightingCrossover.CROSSOVER_TYPES, WeightingCrossover.UNIFORM));
	return types;
    }
}