praktikum-holons/src/exampleAlgorithms/AcoAlgorithm2.java

package exampleAlgorithms;

import java.util.ArrayList;
import java.util.List;
import java.util.ListIterator;

import javax.swing.JFrame;
import javax.swing.JPanel;

import api.AlgorithmFramework;
import ui.model.DecoratedState;

public class AcoAlgorithm2 extends AlgorithmFramework{
	
	//Parameter for Algo with default Values:
	/**
	 * Should be even.
	 */
	private int popsize = 20;
	private int maxGenerations = 200;
	/**
	 * The vaporization factor;
	 */
	private double p = 0.3;
	private double convergenceFactorReset = 0.99;
	private boolean moreInformation = false;
	
	
	
	public AcoAlgorithm2() {
		super();
		addIntParameter("Population", popsize, intValue -> popsize = intValue, 1);
		addIntParameter("maxGenerations", maxGenerations, intValue -> maxGenerations = intValue, 1);
		addDoubleParameter("Vaporization", p, doubleValue -> p = doubleValue, 0.0, 1.0);
		addDoubleParameter("FactorReset", convergenceFactorReset, doubleValue -> convergenceFactorReset = doubleValue, 0.0, 1.0);
		addBooleanParameter("moreInformation", moreInformation, booleanValue -> moreInformation = booleanValue);
	}
	
	
	
	public static void main(String[] args)
	{
	      JFrame newFrame = new JFrame("exampleWindow");
	      AcoAlgorithm2 instance = new AcoAlgorithm2();
	      newFrame.setContentPane(instance.getPanel());
	      newFrame.pack();
	      newFrame.setVisible(true);
	      newFrame.setDefaultCloseOperation(JFrame.EXIT_ON_CLOSE);
	}

	
	
	
	@Override
	protected double evaluateState(DecoratedState actualstate) {
		return Evaluation.getFitnessValueForState(actualstate);
	}


	@Override
	protected int getProgressBarMaxCount() {
		return rounds * popsize * maxGenerations;
	}
	
	
	/** 	
	 * 	Algorithm 20 !! Fitness is better when smaller.: 
	 *  PseudoCode:
	 *  Best <- actual;
	 *  pheromones = initPheromons();
	 * 	for(maxGeneration times){
	 * 		population = createSolutionsBiasedBy(pheromones);
	 * 		for(each Individual i from population){
	 * 			fitness <- evaluatePosition(i);
	 * 			if(fitness < best.fitnessValue) Best <- i;
	 * 		}
	 * 		vaporizeIntensifiePheromons(pheromones);
	 * 	}
	 * @return 
	 */
	@Override
	protected Individual executeAlgo() {
		Individual best = new Individual();
		best.position = extractPositionAndAccess();
		if(moreInformation )console.println("Bit-Array_length: " + best.position.size());
		best.fitness = evaluatePosition(best.position);
		List<Double> runList = new ArrayList<Double>();
		runList.add(best.fitness);
		console.print("Start with: " + best.fitness);
		if(moreInformation)console.println("");
		int problemSize = best.position.size();
		if(problemSize == 0) return best;
		List<Double> pheromones = initPheromones(problemSize);
		List<Individual> population = new ArrayList<Individual>();
		if(moreInformation)console.println("Size To Test:" + population.size());
		for(int generation = 0; generation< maxGenerations; generation++) {
			population.clear();
			population = constructSolutionsBiasedBy(pheromones);
			if(moreInformation)console.println("Generation" + generation + " start with Fitness: " + best.fitness);
			for(Individual i : population) {
				i.fitness = evaluatePosition(i.position);
				if(moreInformation)console.println("Fitness" + i.fitness);
				if(i.fitness < best.fitness) best = i;	
			}
			runList.add(best.fitness);
			if(moreInformation)console.println("________________");
			vaporizeIntensifiePheromons(pheromones, best.position, problemSize);
			double cf = calculateConvergenceFactor(pheromones, problemSize);
			if(moreInformation)console.println("ConvergenceFactor = " + cf);
			if(cf > this.convergenceFactorReset) {
				pheromones = initPheromones(problemSize);
			}
			if(cancel)return null;
		}
		
		
		console.println("   End With:" + best.fitness);
		db.insertNewRun(runList);
		return best;
		
	}
			
			
	/**
	 * tj1 is the pheromon level in the j position
	 * cf is the convergence factor cf e [0;1]
	 * difference = | tj1 - tj0 | = | tj1 - (1 - tj1) |
	 * 
	 * 
	 * 
	 * @param pheromones
	 * @return cf
	 */
	private double calculateConvergenceFactor(List<Double> pheromones,int problemSize) {
		double sumOfDifference = pheromones.stream().map(tj1 -> Math.abs(tj1 - (1.0 - tj1))).reduce(0.0, Double::sum);
		double cf = sumOfDifference / (double)problemSize;
		return cf;
	}
	/**
	 * pheromone <- (1-p) * pheromone;
	 * if(best is true at this position) pheromone <- pheromone + p;
	 * @param pheromones
	 * @param position
	 */
	private void vaporizeIntensifiePheromons(List<Double> pheromones, List<Boolean> position, int problemSize) {
		ListIterator<Double> iterPheromone = pheromones.listIterator();
		ListIterator<Boolean> iterBest = position.listIterator();
		for(int i = 0; i < problemSize; i++) {
			double pheromone = iterPheromone.next();
			boolean bestDecision = iterBest.next();
			iterPheromone.set((1.0 - p) * pheromone + (bestDecision?p:0.0));
		}
	}
	/**
	 * 
	 * @param pheromones
	 * @return
	 */
	private List<Individual> constructSolutionsBiasedBy(List<Double> pheromones) {
		List<Individual> population =  new ArrayList<Individual>();
		for(int i = 0; i < popsize; i++) {
			population.add(constructASolutionBiasedBy(pheromones));
		}
		return population;
	}
	
	
	/**
	 * Walks the path with a ant and decide by pheromones if should take true or false;
	 * A pheromone have a level of 0 < pheromone < 1.
	 * A Pheromone is  equal to the probability.
	 * @param pheromones
	 * @return
	 */
	private Individual constructASolutionBiasedBy(List<Double> pheromones) {
		Individual result = new Individual();
		result.position = new ArrayList<Boolean>();
		for(double pheromone : pheromones) {
			result.position.add((Random.nextDouble() < pheromone));
		}
		return result;
	}
	/**
	 * Initialize Pheromons with 0.5
	 */
	private List<Double> initPheromones(int problemSize) {
		List<Double> result = new ArrayList<Double>();
		for(int i = 0; i < problemSize;i++) {
			result.add(0.5);
		}
		return result;
	}

	
}