public static double Simulate(Specimen spec)
{
var fkf = spec.ToFedKf();
var meas = new double[4];
var n = DirectCosinsMatrix;
//var nt = n.Transpose();
//var c = SensorsOutputCovariances.Inverse();
//var pseudoInverse = (nt * n).Inverse() * nt;
var err = 0.0;
int lng = Math.Min(Signals.RowCount, MaxSimLength);
var results = new Vector3[lng];
results[0] = new Vector3 { X = 0.0, Y = 0.0, Z = 0.0 };
for (int i = 0; i < lng; i++)
{
var sigRow = Signals.Row(i);
var noiseRow = Noises.Row(i);
var targRow = Targets.Row(i);
meas[0] = sigRow[0] + noiseRow[0];
meas[1] = sigRow[1] + noiseRow[1];
meas[2] = sigRow[2] + noiseRow[2];
meas[3] = sigRow[3] + noiseRow[3];
DenseMatrix inps;
if (i > 0)
{
inps = n * results[i - 1].ToMatrix();
}
else
{
inps = new DenseMatrix(4, 1, 0.0);
}
var res = fkf.Step(meas, inps.ToColumnWiseArray());
var errs = new double[] { res[0, 0] - targRow[0], res[1, 0] - targRow[1], res[2, 0] - targRow[2] };
err += (errs[0] * errs[0]) + (errs[1] * errs[1]) + (errs[2] * errs[2]);
results[i] = new Vector3 { X = res[0, 0], Y = res[1, 0], Z = res[2, 0] };
if (PrintSimResults)
{
Console.WriteLine(res.ToColumnWiseArray().Print());
}
if (double.IsNaN(err))
{
return double.NaN;
}
}
if (PrintSimResults)
{
FileParser.Write3ColonFile(@"Data\Evaluations1000.csv", results);
}
return 1/err*lng;
}
private void Initiation()
{
Console.WriteLine("\n=== Starting Initiation ===");
Console.WriteLine("Population Size: {0};\nGenome Size: {1};\nGene Value Range: [{2}; {3}]\n\n",
this.PopulationSize, this.GenomeSize, Specimen.MinGeneValue, Specimen.MaxGeneValue);
_currGeneration = new List <Specimen>();
var newSpecies = Enumerable.Range(0, PopulationSize).AsParallel().Select(i =>
{
var newSpec = new Specimen
{
Length = this.GenomeSize
};
SpecimenHelper.GenerateGenes(ref newSpec);
var fitness = FitnessFunction(newSpec);
// FIXME: Replace '1e5' magic number with well described constant
newSpec.Fitness = double.IsNaN(fitness) ? 0 : (double.IsInfinity(fitness) ? 1e5 : fitness);
Console.WriteLine("Specimen {1} has Fitness: {0}", newSpec.Fitness, i);
return(newSpec);
}).OrderBy(s => s.Fitness).Take(this.PopulationSize);
_currGeneration = newSpecies.ToList(); // Huge load starts here :)
_fitnessTable = new List <double>();
foreach (var spec in _currGeneration)
{
if (!_fitnessTable.Any())
{
_fitnessTable.Add(spec.Fitness);
}
else
{
_fitnessTable.Add(_fitnessTable.Last() + spec.Fitness);
}
}
TotalFitness = _currGeneration.Sum(spec => spec.Fitness);
var best = _currGeneration.Last();
Console.WriteLine("=== Initiation Result ===\n");
Console.WriteLine("Best Specimen:\n{0}", best.Print());
}
private void ShowBestOfAllTime(List <List <Specimen> > generations, PopulationSetting population)
{
Specimen best = null;
Specimen worst = null;
int bestGeneration = 0;
int worstGeneration = 0;
var bestList = new List <Specimen>();
var tempList = new List <Specimen>();
StringBuilder sb = new StringBuilder();
for (int i = 0; i < generations.Count; i++)
{
tempList = generations[i];
tempList.Sort((a, b) => population.FittingFunction(a).CompareTo(population.FittingFunction(b)));
bestList.Add(tempList[0]);
}
for (int i = 0; i < bestList.Count; i++)
{
best = bestList[0];
worst = bestList[1];
if (population.FittingFunction(bestList[i]) < population.FittingFunction(best))
{
best = bestList[i];
bestGeneration = i;
}
if (population.FittingFunction(bestList[i]) > population.FittingFunction(worst))
{
worst = bestList[i];
worstGeneration = i;
}
}
sb.Append("\n\nВывод\n");
sb.Append("Лучший из всех в поколении №" + (bestGeneration + 1) + ": " + best + "\n Значение функции: " + population.FittingFunction(best) + "\n\n");
sb.Append("Худший из всех в поколении №" + (worstGeneration + 1) + ": " + worst + "\n Значение функции: " + population.FittingFunction(worst) + "\n");
Console.WriteLine(sb);
}
private void Selection()
{
var tempGenerationContainer = new ConcurrentBag <Specimen>();
if (this.Elitism)
{
var elite = _currGeneration.Last();
tempGenerationContainer.Add(elite);
}
for (int i = 0; i < this.PopulationSize / 2.5; i++)
{
int pidx1 = this.PopulationSize - i - 1;
int pidx2 = pidx1;
while (pidx1 == pidx2 || _currGeneration[pidx1].IsSimilar(_currGeneration[pidx2]))
{
pidx2 = RouletteSelection();
}
var parent1 = _currGeneration[pidx1].Mutate();
var parent2 = _currGeneration[pidx2].Mutate();
//Console.WriteLine("Selected Species {0} and {1}", pidx1, pidx2);
var children = Rnd.NextDouble() < this.CrossoverRate ? parent1.Crossover(parent2) : new List <Specimen> {
_currGeneration[pidx1], _currGeneration[pidx2]
};
foreach (var ch in children.AsParallel())
{
if (double.IsNaN(ch.Fitness))
{
var fitness = FitnessFunction(ch);
var newChild = new Specimen
{
Genes = ch.Genes,
Length = ch.Length,
Fitness = double.IsNaN(fitness) ? 0 : (double.IsInfinity(fitness) ? 1e5 : fitness)
};
tempGenerationContainer.Add(newChild);
}
else
{
tempGenerationContainer.Add(ch);
}
}
}
_currGeneration = tempGenerationContainer.OrderByDescending(s => s.Fitness).Take(this.PopulationSize).Reverse().ToList();
_fitnessTable = new List <double>();
foreach (var spec in _currGeneration)
{
if (!_fitnessTable.Any())
{
_fitnessTable.Add(spec.Fitness);
}
else
{
_fitnessTable.Add(_fitnessTable.Last() + spec.Fitness);
}
}
TotalFitness = _currGeneration.Sum(spec => spec.Fitness);
Console.WriteLine("=== Selection Result ===\n\n");
Console.WriteLine("\n--- Top 5 ---");
var best = _currGeneration.Last();
Console.WriteLine("Best Specimen:\n{0}\n", best.Print());
int j = 1;
foreach (var spec in _currGeneration.AsEnumerable().Reverse().Take(5).Skip(1))
{
Console.WriteLine("Specimen {0} has Fitness: {1}\n", j++, spec.Fitness); // FIXME: Printout Genes
}
Console.WriteLine("Average Fitness: {0}\n", TotalFitness / PopulationSize);
}
static void Main(string[] args)
{
if (args.Length > 0)
{
var cmd = args[0].ToLower(CultureInfo.InvariantCulture);
switch (cmd)
{
case "simulate":
case "simulation":
InitializeSimulator();
FedKfSim.PrintSimResults = true;
var spec = new Specimen();
SpecimenHelper.SetGenes(ref spec, ReadSimulationGenes());
FedKfSim.Simulate(spec);
break;
case "set":
var settingName = args[1];
var settingValue = args[2];
var config = ConfigurationManager.OpenExeConfiguration(ConfigurationUserLevel.None);
config.AppSettings.Settings[settingName].Value = settingValue;
config.Save(ConfigurationSaveMode.Modified);
ConfigurationManager.RefreshSection("appSettings");
Console.WriteLine("'{0}' set to {1}", settingName, settingValue);
break;
case "print":
Console.WriteLine("Current settings:");
foreach (var name in ConfigurationManager.AppSettings.AllKeys.AsParallel())
{
var value = ConfigurationManager.AppSettings[name];
Console.WriteLine("'{0}' => {1}", name, value);
}
break;
case "help":
case "?":
case "-h":
PrintHelp();
break;
default:
Console.Error.WriteLine(string.Format("\nARGUMENT ERROR\n'{0}' is unknown command!\n", cmd));
PrintHelp();
break;
}
}
else
{
InitializeSimulator();
var genCount = int.Parse(ConfigurationManager.AppSettings["GenerationsCount"]);
var popSize = int.Parse(ConfigurationManager.AppSettings["PopulationSize"]);
var crossOver = double.Parse(ConfigurationManager.AppSettings["CrossoverRate"], FileParser.NumberFormat);
var mutRate = double.Parse(ConfigurationManager.AppSettings["MutationRate"], FileParser.NumberFormat);
var maxGeneVal = double.Parse(ConfigurationManager.AppSettings["MaxGeneValue"], FileParser.NumberFormat);
var minGeneVal = double.Parse(ConfigurationManager.AppSettings["MinGeneValue"], FileParser.NumberFormat);
var genomeLength = int.Parse(ConfigurationManager.AppSettings["GenomeLength"]);
SpecimenHelper.SimilarityThreshold = double.Parse(
ConfigurationManager.AppSettings["SimilarityThreshold"], FileParser.NumberFormat);
var ga = new Ga(genomeLength)
{
FitnessFunction = FedKfSim.Simulate,
Elitism = true,
GenerationsCount = genCount,
PopulationSize = popSize,
CrossoverRate = crossOver,
MutationRate = mutRate
};
FedKfSim.PrintSimResults = false;
ga.Go(maxGeneVal, minGeneVal);
}
Console.ReadLine();
}
private void Selection()
{
var tempGenerationContainer = new ConcurrentBag<Specimen>();
if (this.Elitism)
{
var elite = _currGeneration.Last();
tempGenerationContainer.Add(elite);
}
for (int i = 0; i < this.PopulationSize / 2.5; i++)
{
int pidx1 = this.PopulationSize - i - 1;
int pidx2 = pidx1;
while (pidx1 == pidx2 || _currGeneration[pidx1].IsSimilar(_currGeneration[pidx2]))
{
pidx2 = RouletteSelection();
}
var parent1 = _currGeneration[pidx1].Mutate();
var parent2 = _currGeneration[pidx2].Mutate();
//Console.WriteLine("Selected Species {0} and {1}", pidx1, pidx2);
var children = Rnd.NextDouble() < this.CrossoverRate ? parent1.Crossover(parent2) : new List<Specimen> { _currGeneration[pidx1], _currGeneration[pidx2] };
foreach (var ch in children.AsParallel())
{
if (double.IsNaN(ch.Fitness))
{
var fitness = FitnessFunction(ch);
var newChild = new Specimen
{
Genes = ch.Genes,
Length = ch.Length,
Fitness = double.IsNaN(fitness) ? 0 : (double.IsInfinity(fitness) ? 1e5 : fitness)
};
tempGenerationContainer.Add(newChild);
}
else
{
tempGenerationContainer.Add(ch);
}
}
}
_currGeneration = tempGenerationContainer.OrderByDescending(s => s.Fitness).Take(this.PopulationSize).Reverse().ToList();
_fitnessTable = new List<double>();
foreach (var spec in _currGeneration)
{
if (!_fitnessTable.Any())
{
_fitnessTable.Add(spec.Fitness);
}
else
{
_fitnessTable.Add(_fitnessTable.Last() + spec.Fitness);
}
}
TotalFitness = _currGeneration.Sum(spec => spec.Fitness);
Console.WriteLine("=== Selection Result ===\n\n");
Console.WriteLine("\n--- Top 5 ---");
var best = _currGeneration.Last();
Console.WriteLine("Best Specimen:\n{0}\n", best.Print());
int j = 1;
foreach (var spec in _currGeneration.AsEnumerable().Reverse().Take(5).Skip(1))
{
Console.WriteLine("Specimen {0} has Fitness: {1}\n", j++, spec.Fitness); // FIXME: Printout Genes
}
Console.WriteLine("Average Fitness: {0}\n", TotalFitness/PopulationSize);
}
private void Initiation()
{
Console.WriteLine("\n=== Starting Initiation ===");
Console.WriteLine("Population Size: {0};\nGenome Size: {1};\nGene Value Range: [{2}; {3}]\n\n",
this.PopulationSize, this.GenomeSize, Specimen.MinGeneValue, Specimen.MaxGeneValue);
_currGeneration = new List<Specimen>();
var newSpecies = Enumerable.Range(0, PopulationSize).AsParallel().Select(i =>
{
var newSpec = new Specimen
{
Length = this.GenomeSize
};
SpecimenHelper.GenerateGenes(ref newSpec);
var fitness = FitnessFunction(newSpec);
// FIXME: Replace '1e5' magic number with well described constant
newSpec.Fitness = double.IsNaN(fitness) ? 0 : (double.IsInfinity(fitness) ? 1e5 : fitness);
Console.WriteLine("Specimen {1} has Fitness: {0}", newSpec.Fitness, i);
return newSpec;
}).OrderBy(s => s.Fitness).Take(this.PopulationSize);
_currGeneration = newSpecies.ToList(); // Huge load starts here :)
_fitnessTable = new List<double>();
foreach (var spec in _currGeneration)
{
if (!_fitnessTable.Any())
{
_fitnessTable.Add(spec.Fitness);
}
else
{
_fitnessTable.Add(_fitnessTable.Last() + spec.Fitness);
}
}
TotalFitness = _currGeneration.Sum(spec => spec.Fitness);
var best = _currGeneration.Last();
Console.WriteLine("=== Initiation Result ===\n");
Console.WriteLine("Best Specimen:\n{0}", best.Print());
}
abstract public Specimen CrossoverWith(Specimen secondParent);
public override Specimen CrossoverWith(Specimen secondParent)
{
return(new ArithmeticSpecimen(this, secondParent as ArithmeticSpecimen));
}
