You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

242 lines
5.9 KiB
Java

import javax.swing.*;
import java.awt.*;
import java.util.*;
import java.util.List;
public class GeneticAlgorithm extends JFrame {
Random rnd = new Random(1);
int n = rnd.nextInt(300) + 250;
int generation;
double[] x = new double[n];
double[] y = new double[n];
int[] bestState;
{
for (int i = 0; i < n; i++) {
x[i] = rnd.nextDouble();
y[i] = rnd.nextDouble();
}
}
public void geneticAlgorithm() {
bestState = new int[n];
for (int i = 0; i < n; i++)
bestState[i] = i;
final int populationLimit = 100;
final Population population = new Population(populationLimit);
final int n = x.length;
for (int i = 0; i < populationLimit; i++)
population.chromosomes.add(new Chromosome(optimize(getRandomPermutation(n))));
final double mutationRate = 0.3;
final int generations = 10_000;
for (generation = 0; generation < generations; generation++) {
int i = 0;
while (population.chromosomes.size() < population.populationLimit) {
int i1 = rnd.nextInt(population.chromosomes.size());
int i2 = (i1 + 1 + rnd.nextInt(population.chromosomes.size() - 1)) % population.chromosomes.size();
Chromosome parent1 = population.chromosomes.get(i1);
Chromosome parent2 = population.chromosomes.get(i2);
int[][] pair = crossOver(parent1.p, parent2.p);
if (rnd.nextDouble() < mutationRate) {
mutate(pair[0]);
mutate(pair[1]);
}
population.chromosomes.add(new Chromosome(optimize(pair[0])));
population.chromosomes.add(new Chromosome(optimize(pair[1])));
}
population.nextGeneration();
bestState = population.chromosomes.get(0).p;
repaint();
}
}
int[][] crossOver(int[] p1, int[] p2) {
int n = p1.length;
int i1 = rnd.nextInt(n);
int i2 = (i1 + 1 + rnd.nextInt(n - 1)) % n;
int[] n1 = p1.clone();
int[] n2 = p2.clone();
boolean[] used1 = new boolean[n];
boolean[] used2 = new boolean[n];
for (int i = i1; ; i = (i + 1) % n) {
n1[i] = p2[i];
used1[n1[i]] = true;
n2[i] = p1[i];
used2[n2[i]] = true;
if (i == i2) {
break;
}
}
for (int i = (i2 + 1) % n; i != i1; i = (i + 1) % n) {
if (used1[n1[i]]) {
n1[i] = -1;
} else {
used1[n1[i]] = true;
}
if (used2[n2[i]]) {
n2[i] = -1;
} else {
used2[n2[i]] = true;
}
}
int pos1 = 0;
int pos2 = 0;
for (int i = 0; i < n; i++) {
if (n1[i] == -1) {
while (used1[pos1])
++pos1;
n1[i] = pos1++;
}
if (n2[i] == -1) {
while (used2[pos2])
++pos2;
n2[i] = pos2++;
}
}
return new int[][]{n1, n2};
}
void mutate(int[] p) {
int n = p.length;
int i = rnd.nextInt(n);
int j = (i + 1 + rnd.nextInt(n - 1)) % n;
reverse(p, i, j);
}
// http://en.wikipedia.org/wiki/2-opt
static void reverse(int[] p, int i, int j) {
int n = p.length;
// reverse order from i to j
while (i != j) {
int t = p[j];
p[j] = p[i];
p[i] = t;
i = (i + 1) % n;
if (i == j) break;
j = (j - 1 + n) % n;
}
}
double eval(int[] state) {
double res = 0;
for (int i = 0, j = state.length - 1; i < state.length; j = i++)
res += dist(x[state[i]], y[state[i]], x[state[j]], y[state[j]]);
return res;
}
static double dist(double x1, double y1, double x2, double y2) {
double dx = x1 - x2;
double dy = y1 - y2;
return Math.sqrt(dx * dx + dy * dy);
}
int[] getRandomPermutation(int n) {
int[] res = new int[n];
for (int i = 0; i < n; i++) {
int j = rnd.nextInt(i + 1);
res[i] = res[j];
res[j] = i;
}
return res;
}
// try all 2-opt moves
int[] optimize(int[] p) {
int[] res = p.clone();
for (boolean improved = true; improved; ) {
improved = false;
for (int i = 0; i < n; i++) {
for (int j = 0; j < n; j++) {
if (i == j || (j + 1) % n == i) continue;
int i1 = (i - 1 + n) % n;
int j1 = (j + 1) % n;
double delta = dist(x[res[i1]], y[res[i1]], x[res[j]], y[res[j]])
+ dist(x[res[i]], y[res[i]], x[res[j1]], y[res[j1]])
- dist(x[res[i1]], y[res[i1]], x[res[i]], y[res[i]])
- dist(x[res[j]], y[res[j]], x[res[j1]], y[res[j1]]);
if (delta < -1e-9) {
reverse(res, i, j);
improved = true;
}
}
}
}
return res;
}
class Chromosome implements Comparable<Chromosome> {
final int[] p;
private double cost = Double.NaN;
public Chromosome(int[] p) {
this.p = p;
}
public double getCost() {
return Double.isNaN(cost) ? cost = eval(p) : cost;
}
@Override
public int compareTo(Chromosome o) {
return Double.compare(getCost(), o.getCost());
}
}
static class Population {
List<Chromosome> chromosomes = new ArrayList<>();
final int populationLimit;
public Population(int populationLimit) {
this.populationLimit = populationLimit;
}
public void nextGeneration() {
Collections.sort(chromosomes);
chromosomes = new ArrayList<>(chromosomes.subList(0, (chromosomes.size() + 1) / 2));
}
}
// visualization code
public GeneticAlgorithm() {
setContentPane(new JPanel() {
protected void paintComponent(Graphics g) {
super.paintComponent(g);
((Graphics2D) g).setRenderingHint(RenderingHints.KEY_ANTIALIASING, RenderingHints.VALUE_ANTIALIAS_ON);
((Graphics2D) g).setStroke(new BasicStroke(3));
g.setColor(Color.BLUE);
int w = getWidth() - 5;
int h = getHeight() - 30;
for (int i = 0, j = n - 1; i < n; j = i++)
g.drawLine((int) (x[bestState[i]] * w), (int) ((1 - y[bestState[i]]) * h),
(int) (x[bestState[j]] * w), (int) ((1 - y[bestState[j]]) * h));
g.setColor(Color.RED);
for (int i = 0; i < n; i++)
g.drawOval((int) (x[i] * w) - 1, (int) ((1 - y[i]) * h) - 1, 3, 3);
g.setColor(Color.BLACK);
g.drawString(String.format("length: %.3f", eval(bestState)), 5, h + 20);
g.drawString(String.format("generation: %d", generation), 150, h + 20);
}
});
setSize(new Dimension(600, 600));
setDefaultCloseOperation(WindowConstants.EXIT_ON_CLOSE);
setVisible(true);
new Thread(this::geneticAlgorithm).start();
}
public static void main(String[] args) {
new GeneticAlgorithm();
}
}