Обучить перцептрон с помощью генетического алгоритма

public class ChromoPerceptron {

double[] enters; double outer; double[] weights; double[][] patterns = { {0, 0}, {1, 0}, {0, 1}, {1, 1}

}; double fitness; double answers[] = {0, 1, 1, 1};

ChromoPerceptron(){ enters = new double[2]; weights = new double[enters.length]; // for (int i = 0; i < weights.length; i++) { // weights[i] = 0.2 * Math.random() + 0.1; // } }

void countFitness() { fitness = 0; for (int p = 0; p < patterns.length; p++) { enters = java.util.Arrays.copyOf(patterns[p], patterns[p].length); countOuter(); double error = answers[p] - outer; fitness += Math.abs(error); // for (int i = 0; i < enters.length; i++) { // weights[i] += 0.1 error enters[i]; // } } }

public void countOuter() { outer = 0; for (int i = 0; i < enters.length; i++) { outer += enters[i] * weights[i]; } if (outer > 0.5) outer = 1; else outer = 0; }

public void test() { for (int p = 0; p < patterns.length; p++) { enters = java.util.Arrays.copyOf(patterns[p], patterns[p].length); countOuter(); System.out.println(outer);

}
System.out.println(weights[0]+" "+weights[1]);

} }

import java.util.Comparator; import java.util.ArrayList; import java.util.Comparator;

public class GenMultPercetron { java.util.List proba; java.util.List chromosomes; static int POPULATION_SIZE = 1000; GenMultPercetron(){ init(); for(int i=0; i<1000; i++){ mark(); selection(); //selectionRulet(); recombination();

    }
    mark();
    selection();
    //selectionRulet();
    ChromoPerceptron best = chromosomes.get(0);
    //System.out.println(best.weights[0]+" "+best.weights[1]+" "+best.fitness);
    best.test();
}

void init(){
    chromosomes = new java.util.ArrayList<>();
    proba = new java.util.ArrayList<>();
    java.util.Random r = new java.util.Random();
    for(int i=0; i<100; i++){
        ChromoPerceptron chromosome = new ChromoPerceptron();
        chromosome.weights[0] = r.nextDouble();
        chromosome.weights[1] = r.nextDouble();
        chromosomes.add(chromosome);
    }
}

void mark(){
    for(ChromoPerceptron chromosome:chromosomes){
        chromosome.countFitness();
    }
}

void selection(){
    chromosomes.sort(new Comparator(){
        @Override
        public int compare(Object obj1, Object obj2){
            int res = 0;
            double fitness1 = ((ChromoPerceptron)obj1).fitness;
            double fitness2 = ((ChromoPerceptron)obj2).fitness;
            if (fitness2<fitness1) res=1; else
            if(fitness2>fitness1) res = -1; else res=0;
            return res;
        }
    });

    chromosomes = chromosomes.subList(0, chromosomes.size()/2);
    for(ChromoPerceptron chromosome:chromosomes){
        System.out.println(chromosome.weights[0]+" "+chromosome.weights[1]+" "+chromosome.fitness);;
    }
    System.out.println("__________________________________________________________");
}

void selection1() { chromosomes.sort(new Comparator() { @Override public int compare(Object obj1, Object obj2) { int res = 0; double prob1 = ((Chromosome) obj1).selectionProb; double prob2 = ((Chromosome) obj2).selectionProb; if (prob2 < prob1) res = 1; else if (prob2 > prob1) res = -1; else res = 0; return res; } }); }

void recombination(){
    java.util.Random r = new java.util.Random();
    java.util.ArrayList<ChromoPerceptron> chromosomes2 = new java.util.ArrayList<>();
    for(int i=0; i<50; i++){
        int index = r.nextInt(chromosomes.size());
        ChromoPerceptron c1 = chromosomes.get(index);

        index = r.nextInt(chromosomes.size());
        ChromoPerceptron c2 = chromosomes.get(index);

        ChromoPerceptron c1new = new ChromoPerceptron();
        c1new.weights[0] = c1.weights[0];
        c1new.weights[1] = c2.weights[1];

        ChromoPerceptron c2new = new ChromoPerceptron();
        c2new.weights[0] = c2.weights[0];
        c2new.weights[1] = c1.weights[1];

        chromosomes2.add(c1new);
        chromosomes2.add(c2new);
    }
    chromosomes = chromosomes2;
}

}

Marfein / -12

Обучить перцептрон с помощью генетического алгоритма #9