Free cookie consent management tool by TermsFeed Policy Generator

source: branches/crossvalidation-2434/HeuristicLab.Tests/HeuristicLab.Scripting-3.3/Script Sources/OSGARastriginScriptSource.cs @ 13878

Last change on this file since 13878 was 11789, checked in by jkarder, 10 years ago

#2262: refactored scripting unit tests

  • changed code template loading
  • minor code changes
File size: 3.6 KB
Line 
1using System;
2using System.Linq;
3
4public class OSGARastriginScript : HeuristicLab.Scripting.CSharpScriptBase {
5  public override void Main() {
6    int N = vars.N.Value;
7    double minX = vars.minX.Value;
8    double maxX = vars.maxX.Value;
9    int seed = vars.seed.Value;
10    var rand = new Random(seed);
11
12    var result = OSGA(rand, popSize: 1000, iterations: 5000, maxSelPres: 1000, mutationRate: 0.1,
13         obj: Rastrigin,
14      // x ~(i.i.d) U(0,1) * 10.24 - 5.12
15         creator: () => Enumerable.Range(1, N).Select(_ => rand.NextDouble() * (maxX - minX) + minX).ToArray(),
16      // random parent selection
17         selector: (f) => rand.Next(f.Length),
18      // single point crossover
19         crossover: (p0, p1, cand) => {
20           int cut = rand.Next(cand.Length);
21           Array.Copy(p0, cand, cut);
22           Array.Copy(p1, cut, cand, cut, cand.Length - cut);
23         },
24      // single point manipulation
25         manipulator: (p) => p[rand.Next(N)] = rand.NextDouble() * (maxX - minX) + minX,
26         output: true
27      );
28
29    double bestFit = result.Item2;
30    vars.bestFitness = bestFit;
31    Console.WriteLine("best fitness: {0}", bestFit);
32  }
33
34  private double Rastrigin(double[] x) {
35    return 10.0 * x.Length + x.Sum(xi => xi * xi - 10.0 * Math.Cos(2.0 * Math.PI * xi));
36  }
37
38  private Tuple<T, double> OSGA<T>(Random rand, int popSize, int iterations, double maxSelPres, double mutationRate, Func<T, double> obj,
39      Func<T> creator, Func<double[], int> selector, Action<T, T, T> crossover, Action<T> manipulator,
40      bool output = false)
41      where T : ICloneable {
42    // generate random pop
43    T[] pop = Enumerable.Range(0, popSize).Select(_ => creator()).ToArray();
44    // evaluate initial pop
45    double[] fit = pop.Select(p => obj(p)).ToArray();
46
47    // arrays for next pop (clone current pop because solutions are reused)
48    T[] popNew = pop.Select(pi => (T)pi.Clone()).ToArray();
49    double[] fitNew = new double[popSize];
50
51    var bestSolution = (T)pop.First().Clone(); // take a random solution (don't care)
52    double bestFit = fit.First();
53
54    // run generations
55    double curSelPres = 0;
56    for (int g = 0; g < iterations && curSelPres < maxSelPres; g++) {
57      // keep the first element as elite
58      int i = 1;
59      int genEvals = 0;
60      do {
61        var p1Idx = selector(fit);
62        var p2Idx = selector(fit);
63
64        var p1 = pop[p1Idx];
65        var p2 = pop[p2Idx];
66        var f1 = fit[p1Idx];
67        var f2 = fit[p2Idx];
68
69        // generate candidate solution (reuse old solutions)
70        crossover(p1, p2, popNew[i]);
71        // optional mutation
72        if (rand.NextDouble() < mutationRate) {
73          manipulator(popNew[i]);
74        }
75
76        double f = obj(popNew[i]);
77        genEvals++;
78        // if child is better than best (strict offspring selection)
79        if (f < Math.Min(f1, f2)) {
80          // update best fitness
81          if (f < bestFit) {
82            bestFit = f;
83            bestSolution = (T)popNew[i].Clone(); // overall best
84          }
85
86          // keep
87          fitNew[i] = f;
88          i++;
89        }
90
91        curSelPres = genEvals / (double)popSize;
92      } while (i < popNew.Length && curSelPres < maxSelPres);
93
94      Console.WriteLine("generation {0} obj {1:0.000} sel. pres. {2:###.0}", g, bestFit, curSelPres);
95
96      // swap
97      var tmpPop = pop;
98      var tmpFit = fit;
99      pop = popNew;
100      fit = fitNew;
101      popNew = tmpPop;
102      fitNew = tmpFit;
103
104      // keep elite
105      popNew[0] = (T)bestSolution.Clone();
106      fitNew[0] = bestFit;
107    }
108
109    return Tuple.Create(bestSolution, bestFit);
110  }
111}
Note: See TracBrowser for help on using the repository browser.