Free cookie consent management tool by TermsFeed Policy Generator

source: trunk/HeuristicLab.Problems.Instances.DataAnalysis/3.3/Regression/Feynman/Feynman90.cs @ 17945

Last change on this file since 17945 was 17805, checked in by gkronber, 4 years ago

#3075 Use the same noise levels and calculation as in our experiments for the IEEE TeC paper. Reordered instances by name first and noise level second. Removed number of samples from the name.

File size: 3.6 KB
RevLine 
[17647]1using System;
2using System.Collections.Generic;
3using System.Linq;
4using HeuristicLab.Common;
5using HeuristicLab.Random;
6
7namespace HeuristicLab.Problems.Instances.DataAnalysis {
8  public class Feynman90 : FeynmanDescriptor {
9    private readonly int testSamples;
10    private readonly int trainingSamples;
11
12    public Feynman90() : this((int) DateTime.Now.Ticks, 10000, 10000, null) { }
13
14    public Feynman90(int seed) {
15      Seed            = seed;
16      trainingSamples = 10000;
17      testSamples     = 10000;
18      noiseRatio      = null;
19    }
20
21    public Feynman90(int seed, int trainingSamples, int testSamples, double? noiseRatio) {
22      Seed                 = seed;
23      this.trainingSamples = trainingSamples;
24      this.testSamples     = testSamples;
25      this.noiseRatio      = noiseRatio;
26    }
27
28    public override string Name {
29      get {
30        return string.Format(
[17805]31          "III.9.52 (p_d*Ef*t/h*sin((omega-omega_0)*t/2)**2/((omega-omega_0)*t/2)**2 | {0}",
32           noiseRatio == null ? "no noise" : string.Format(System.Globalization.CultureInfo.InvariantCulture, "noise={0:g}",noiseRatio));
[17647]33      }
34    }
35
36    protected override string TargetVariable { get { return noiseRatio == null ? "prob" : "prob_noise"; } }
37
38    protected override string[] VariableNames {
39      get { return new[] {"p_d", "Ef", "t", "h", "omega", "omega_0", noiseRatio == null ? "prob" : "prob_noise"}; }
40    }
41
42    protected override string[] AllowedInputVariables {
43      get { return new[] {"p_d", "Ef", "t", "h", "omega", "omega_0"}; }
44    }
45
46    public int Seed { get; private set; }
47
48    protected override int TrainingPartitionStart { get { return 0; } }
49    protected override int TrainingPartitionEnd { get { return trainingSamples; } }
50    protected override int TestPartitionStart { get { return trainingSamples; } }
51    protected override int TestPartitionEnd { get { return trainingSamples + testSamples; } }
52
53    protected override List<List<double>> GenerateValues() {
54      var rand = new MersenneTwister((uint) Seed);
55
56      var data    = new List<List<double>>();
57      var p_d     = ValueGenerator.GenerateUniformDistributedValues(rand.Next(), TestPartitionEnd, 1, 3).ToList();
58      var Ef      = ValueGenerator.GenerateUniformDistributedValues(rand.Next(), TestPartitionEnd, 1, 3).ToList();
59      var t       = ValueGenerator.GenerateUniformDistributedValues(rand.Next(), TestPartitionEnd, 1, 3).ToList();
60      var h       = ValueGenerator.GenerateUniformDistributedValues(rand.Next(), TestPartitionEnd, 1, 3).ToList();
61      var omega   = ValueGenerator.GenerateUniformDistributedValues(rand.Next(), TestPartitionEnd, 1, 5).ToList();
62      var omega_0 = ValueGenerator.GenerateUniformDistributedValues(rand.Next(), TestPartitionEnd, 1, 5).ToList();
63
64      var prob = new List<double>();
65
66      data.Add(p_d);
67      data.Add(Ef);
68      data.Add(t);
69      data.Add(h);
70      data.Add(omega);
71      data.Add(omega_0);
72      data.Add(prob);
73
74      for (var i = 0; i < p_d.Count; i++) {
75        var res = p_d[i] * Ef[i] * t[i] / h[i] *
76                  Math.Pow(Math.Sin((omega[i] - omega_0[i]) * t[i] / 2), 2) /
77                  Math.Pow((omega[i] - omega_0[i]) * t[i] / 2, 2);
78        prob.Add(res);
79      }
80
81      if (noiseRatio != null) {
82        var prob_noise  = new List<double>();
[17805]83        var sigma_noise = (double) Math.Sqrt(noiseRatio.Value) * prob.StandardDeviationPop();
[17647]84        prob_noise.AddRange(prob.Select(md => md + NormalDistributedRandom.NextDouble(rand, 0, sigma_noise)));
85        data.Remove(prob);
86        data.Add(prob_noise);
87      }
88
89      return data;
90    }
91  }
92}
Note: See TracBrowser for help on using the repository browser.