GA.java

/***********************************************************************

	This file is part of KEEL-software, the Data Mining tool for regression, 
	classification, clustering, pattern mining and so on.

	Copyright (C) 2004-2010
	
	F. Herrera (herrera@decsai.ugr.es)
    L. S�nchez (luciano@uniovi.es)
    J. Alcal�-Fdez (jalcala@decsai.ugr.es)
    S. Garc�a (sglopez@ujaen.es)
    A. Fern�ndez (alberto.fernandez@ujaen.es)
    J. Luengo (julianlm@decsai.ugr.es)

	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 3 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, see http://www.gnu.org/licenses/
  
**********************************************************************/

/**
 * <p>
 * @author Written by Jaume Bacardit (La Salle, Ram�n Llull University - Barcelona) 28/03/2004
 * @author Modified by Xavi Sol� (La Salle, Ram�n Llull University - Barcelona) 23/12/2008
 * @author Modified by Jose A. Saez Munoz (ETSIIT, Universidad de Granada - Granada) 10/09/10
 * @version 1.1
 * @since JDK1.2
 * </p>
 */


package keel.Algorithms.Genetic_Rule_Learning.MPLCS;

import java.util.*;
import keel.Algorithms.Genetic_Rule_Learning.MPLCS.Assistant.Globals.*;


public class GA {

  Classifier[] population;
  Classifier[] bestIndiv;
  int numVersions;

  /** Creates a new instance of GA */
  public GA() {
  }

  /**
   *  Prepares GA for a new run.
   */
  public void initGA() {
    //Init population
    population = new Classifier[Parameters.popSize];
    PopulationWrapper.initInstancesEvaluation();

    numVersions = PopulationWrapper.numVersions();
    bestIndiv = new Classifier[numVersions];

    Factory.initialize();
    initPopulation(population);
    Statistics.initStatistics();
  }

  /**
   *  Inits a new population.
   *  @param _population Population to init.
   */
  private void initPopulation(Classifier[] _population) {
    for (int i = 0; i < Parameters.popSize; i++) {
      _population[i] = Factory.newClassifier();
      _population[i].initRandomClassifier();
    }
  }

  public void checkBestIndividual() {
    Classifier best = PopulationWrapper.getBest(population);
    int currVer = PopulationWrapper.getCurrentVersion();

    if (bestIndiv[currVer] == null) {
      bestIndiv[currVer] = best.copy();
    }
    else {
      if (best.compareToIndividual(bestIndiv[currVer])) {
        bestIndiv[currVer] = best.copy();
      }
    }
  }

  /**
   *  Executes a number of iterations of GA.
   */
  public void run() {
    Classifier[] offsprings;

    PopulationWrapper.doEvaluation(population);

    int numIterations = Parameters.numIterations;

    for (int iteration = 0; iteration < numIterations; iteration++) {
    	    	
      boolean lastIteration = (iteration == numIterations - 1);
      Parameters.percentageOfLearning = (double) iteration
          / (double) numIterations;
      boolean res1 = PopulationWrapper.initIteration();
      boolean res2 = Timers.runTimers(iteration, population);
      if (res1 || res2) {
        PopulationWrapper.setModified(population);
      }

      // GA cycle
      population = doTournamentSelection(population);
      offsprings = doCrossover(population);
      doMutation(offsprings);
      
      if (!Factory.realKR){
    	  localSearch(offsprings);
      }
	  
      PopulationWrapper.doEvaluation(offsprings);
      population = replacementPolicy(offsprings, lastIteration);

      Statistics.computeStatistics(population);
      Timers.runOutputTimers(iteration, population);
    }

    Statistics.statisticsToFile();
    Classifier best = PopulationWrapper.getBest(population);
    
	Main.numRulesBest = best.numRules;
    if(Globals_DefaultC.enabled)
    	Main.numRulesBest++;


    LogManager.println("\nPhenotype: ");
    best.printClassifier();
    PopulationWrapper.testClassifier(best,"training",Parameters.train2InputFile,Parameters.trainOutputFile);
    PopulationWrapper.testClassifier(best,"test",Parameters.testInputFile,Parameters.testOutputFile);
  }
  
    
  public void localSearch(Classifier[] _population){
	  
	    int popSize = Parameters.popSize;
	    double probLocalSearch = Parameters.probLocalSearch;

	    for (int i = 0; i < popSize ; ++i){
	      if (Rand.getReal() < probLocalSearch) {
	        _population[i].doLocalSearch();
	      }
	    }
	  
  }
  

  
  	Classifier[] doCrossover(Classifier[] _population){
  		
  		Chronometer.startChronCrossover();

  	    int i, j, k, countCross = 0;
  	    int numNiches = _population[0].getNumNiches();
  	    
  	    ArrayList[] parents = new ArrayList[numNiches];
  	    Classifier parent1, parent2;
  	    Classifier[] offsprings = new Classifier[2];
  	    Classifier[] offspringPopulation = new Classifier[Parameters.popSize];

  	    for (i = 0; i < numNiches; i++) {
  	    	parents[i] = new ArrayList();
  	    	parents[i].ensureCapacity(Parameters.popSize);
  	    }

  	    for (i = 0; i < Parameters.popSize; i++) {
  	    	int niche = _population[i].getNiche();
  	    	parents[niche].add(new Integer(i));
  	    }

  	    for (i = 0; i < numNiches; i++) {
  	    	
  	    	int size = parents[i].size();
  	  	  	Sampling samp = new Sampling(size);
  	  	  	int p1 = -1;
  	      
  	  	  	for (j = 0; j < size; j++) {

  	  	  		if (Rand.getReal() < Parameters.probCrossover) {
  	        	
  	  	  			if (Rand.getReal() < Parameters.probRSWcrossover){
  					
  	  	  				int num = Parameters.numParentsRSWcrossover;	
  	  	  				int posParents[] = new int[num];
  					
  	  	  				for (k = 0; k < num; k++)
  	  	  					posParents[k] = (Integer)parents[i].get(samp.getSample());
  	  	  				
  	  	  				offspringPopulation[countCross++] = RSW(_population, posParents, num).copy();
  	  	  			}

  	  	  			else{
  	  	  				  	  	  				
  	  	  				if (p1 == -1) {    	  	            
  	  	  					p1 = samp.getSample();
  	  	  				}

  	  	  				else {
  	  	  					int p2 = samp.getSample();
  	  	  					int pos1 = ( (Integer) parents[i].get(p1)).intValue();
  	  	  					int pos2 = ( (Integer) parents[i].get(p2)).intValue();
  	  	  					parent1 = _population[pos1];
  	  	  					parent2 = _population[pos2];


  	  	  					offsprings = parent1.crossoverClassifiers(parent2);
  	  	  					offspringPopulation[countCross++] = offsprings[0];
  	  	  					offspringPopulation[countCross++] = offsprings[1];
  	  	  					p1 = -1;
  	  	  				}	
  	  	  			}

  	  	  		}

  	        
  	  	  		else {
  	  	  			int pos = ( (Integer) parents[i].get(samp.getSample())).intValue();
  	  	  			offspringPopulation[countCross++] = _population[pos].copy();
  	  	  		}
  	      
  	  	  	}


  	  	  	if (p1 != -1) {
  	  	  		int pos = ( (Integer) parents[i].get(p1)).intValue();
  	  	  		offspringPopulation[countCross++] = _population[pos].copy();
  	  	  	}
  	    }

  	    
  	    Chronometer.stopChronCrossover();
  	    return offspringPopulation;
  	}



  	public Classifier RSW(Classifier[] _population, int[] parentsPos,  int numParents){
  		
  		
  		Classifier parents[]=new Classifier[numParents];

  		for (int i = 0; i < numParents; i++)
  			parents[i] = _population[parentsPos[i]].copy();
		
  		Classifier newSon = _population[parentsPos[0]].copy();
  		newSon.crossoverRSW(parents, numParents);
  		
  		return newSon;
  	}  

  	
  private int selectNicheWOR(int[] quotas) {
    int num = quotas.length;
    if (num == 1) {
      return 0;
    }

    int total = 0, i;
    for (i = 0; i < num; i++) {
      total += quotas[i];
    }
    if (total == 0) {
      return Rand.getInteger(0, num - 1);
    }
    int pos = Rand.getInteger(0, total - 1);
    total = 0;
    for (i = 0; i < num; i++) {
      total += quotas[i];
      if (pos < total) {
        quotas[i]--;
        return i;
      }
    }

    LogManager.printErr("We should not be here");
    System.exit(1);
    return -1;
  }

  private void initPool(ArrayList pool, int whichNiche,
                        Classifier[] _population) {
    if (Globals_DefaultC.nichingEnabled) {
      for (int i = 0; i < Parameters.popSize; i++) {
        if (_population[i].getNiche() == whichNiche) {
          pool.add(new Integer(i));
        }
      }
    }
    else {
      for (int i = 0; i < Parameters.popSize; i++) {
        pool.add(new Integer(i));
      }
    }
  }

  private int selectCandidateWOR(ArrayList pool, int whichNiche,
                                 Classifier[] _population) {
    if (pool.size() == 0) {
      initPool(pool, whichNiche, population);
      if (pool.size() == 0) {
        return Rand.getInteger(0, Parameters.popSize - 1);
      }
    }

    int pos = Rand.getInteger(0, pool.size() - 1);
    int elem = ( (Integer) pool.get(pos)).intValue();
    pool.remove(pos);
    return elem;
  }

  /**
   *  Does Tournament Selection without replacement.
   */
  public Classifier[] doTournamentSelection(Classifier[] _population) {
    Chronometer.startChronSelection();

    Classifier[] selectedPopulation;
    selectedPopulation = new Classifier[Parameters.popSize];
    int i, j, winner, candidate;
    int numNiches;
    if (Globals_DefaultC.nichingEnabled) {
      numNiches = _population[0].getNumNiches();
    }
    else {
      numNiches = 1;
    }

    ArrayList[] pools = new ArrayList[numNiches];
    for (i = 0; i < numNiches; i++) {
      pools[i] = new ArrayList();
    }
    int[] nicheCounters = new int[numNiches];
    int nicheQuota = Parameters.popSize / numNiches;
    for (i = 0; i < numNiches; i++) {
      nicheCounters[i] = nicheQuota;
    }

    for (i = 0; i < Parameters.popSize; i++) {
      // There can be only one
      int niche = selectNicheWOR(nicheCounters);
      winner = selectCandidateWOR(pools[niche], niche
                                  , _population);
      for (j = 1; j < Parameters.tournamentSize; j++) {
        candidate = selectCandidateWOR(pools[niche]
                                       , niche, _population);
        if (_population[candidate].compareToIndividual(_population[winner])) {
          winner = candidate;
        }
      }
      selectedPopulation[i] = _population[winner].copy();
    }
    Chronometer.stopChronSelection();
    return selectedPopulation;
  }

  public void doMutation(Classifier[] _population) {
    Chronometer.startChronMutation();
    int popSize = Parameters.popSize;
    double probMut = Parameters.probMutationInd;

    for (int i = 0; i < Parameters.popSize; i++) {
      if (Rand.getReal() < probMut) {
        _population[i].doMutation();
      }
    }

    doSpecialStages(_population);

    Chronometer.stopChronMutation();
  }

  void sortedInsert(ArrayList set, Classifier cl) {
    for (int i = 0, max = set.size(); i < max; i++) {
      if (cl.compareToIndividual( (Classifier) set.get(i))) {
        set.add(i, cl);
        return;
      }
    }
    set.add(cl);
  }

  public Classifier[] replacementPolicy(Classifier[] offspring
                                        , boolean lastIteration) {
    int i;

    Chronometer.startChronReplacement();

    if (lastIteration) {
      for (i = 0; i < numVersions; i++) {
        if (bestIndiv[i] != null) {
          PopulationWrapper.evaluateClassifier(
              bestIndiv[i]);
        }
      }
      ArrayList set = new ArrayList();
      for (i = 0; i < Parameters.popSize; i++) {
        sortedInsert(set, offspring[i]);
      }
      for (i = 0; i < numVersions; i++) {
        if (bestIndiv[i] != null) {
          sortedInsert(set, bestIndiv[i]);
        }
      }

      for (i = 0; i < Parameters.popSize; i++) {
        offspring[i] = (Classifier) set.get(i);
      }
    }
    else {
      boolean previousVerUsed = false;
      int currVer = PopulationWrapper.getCurrentVersion();
      if (bestIndiv[currVer] == null && currVer > 0) {
        previousVerUsed = true;
        currVer--;
      }

      if (bestIndiv[currVer] != null) {
        PopulationWrapper.evaluateClassifier(bestIndiv[currVer]);
        int worst = PopulationWrapper.getWorst(offspring);
        offspring[worst] = bestIndiv[currVer].copy();
      }
      if (!previousVerUsed) {
        int prevVer;
        if (currVer == 0) {
          prevVer = numVersions - 1;
        }
        else {
          prevVer = currVer - 1;
        }
        if (bestIndiv[prevVer] != null) {
          PopulationWrapper.evaluateClassifier(bestIndiv[prevVer]);
          int worst = PopulationWrapper.getWorst(offspring);
          offspring[worst] = bestIndiv[prevVer].copy();
        }
      }
    }

    Chronometer.stopChronReplacement();
    return offspring;
  }

  public void doSpecialStages(Classifier[] population) {
    int numStages = population[0].numSpecialStages();

    for (int i = 0; i < numStages; i++) {
      for (int j = 0; j < population.length; j++) {
        population[j].doSpecialStage(i);
      }
    }
  }
}