private Specimen[] Mutate(Specimen[] population)
{
Specimen[] newPopulation = new Specimen[_populationSize];
int i = 0;
foreach (var t in population)
{
if (_random.Next(101) > _mutationFactor)
{
switch (_mutationMethod)
{
case 1:
newPopulation[i] = InvertMutation(t.Path);
break;
case 2:
newPopulation[i] = ScrumbleMutation(t.Path);
break;
}
}
else
{
newPopulation[i] = t;
}
i++;
}
return(newPopulation);
}
private Specimen[] RateAndSort(Specimen[] population)
{
population = population.OrderBy(x => x.Cost).ToArray <Specimen>(); //tu może będzie trzeba porawić
if (population[0].Cost < _bestSpecimen.Cost)
{
_bestSpecimen = population[0];
}
return(population);
}
public TSPGenetic(int[][] matrix, long ttl, int populationSize, float mutationFactor, float crossingFactor,
int mutationMethod, int crossingMethod)
{
_matrix = matrix;
_ttl = ttl * 1000;
_populationSize = populationSize;
_mutationFactor = 100 - (int)(mutationFactor * 100);
_crossingFactor = 100 - (int)(crossingFactor * 100);
_mutationMethod = mutationMethod;
_crossingMethod = crossingMethod;
_size = matrix.Length;
_random = new Random();
_scrumbleIterations = (int)Math.Sqrt(_size);
_bestSpecimen = new Specimen();
_bestSpecimen.Cost = int.MaxValue;
}
public void run()
{
DateTime end = DateTime.Now.AddMilliseconds(_ttl);
Specimen[] population = new Specimen[_populationSize];
//population[0] = GenerateGreedy();
for (int i = 0; i < _populationSize; i++)
{
population[i] = new Specimen();
population[i].Path = randomPath();
population[i].Cost = rateRute(population[i].Path);
}
population = RateAndSort(population);
while (end.CompareTo(DateTime.Now) > 0)// jak nie działa to tu inny znak
{
population = Crossing(population);
population = Mutate(population);
population = RateAndSort(population);
}
}
private Specimen ScrumbleMutation(int[] parrent1)
{
int[] offspring = (int[])parrent1.Clone();
for (int i = 0; i < _scrumbleIterations; i++)
{
int a, b, t;
do
{
a = _random.Next(_size - 2) + 1;
b = _random.Next(_size - 2) + 1;
} while (a == b);
t = offspring[a];
offspring[a] = offspring[b];
offspring[b] = t;
}
Specimen toReturn = new Specimen();
toReturn.Path = offspring;
toReturn.Cost = rateRute(offspring);
return(toReturn);
}
private Specimen InvertMutation(int[] parrent)
{
int[] offspring = new int[parrent.Length];
int a = 0, b = 0;
do
{
a = _random.Next(parrent.Length - 2) + 1;
b = _random.Next(parrent.Length - 2) + 1;
if (a > b)
{
int t = a;
a = b;
b = t;
}
} while (a == b);
for (int i = 0; i < a; i++)
{
offspring[i] = parrent[i];
}
for (int i = 0; i < b - a + 1; i++)
{
offspring[a + i] = parrent[b - i];
}
for (int i = b + 1; i < parrent.Length; i++)
{
offspring[i] = parrent[i];
}
Specimen toReturn = new Specimen();
toReturn.Path = offspring;
toReturn.Cost = rateRute(offspring);
return(toReturn);
}
private Specimen GenerateGreedy()
{
Specimen toReturn = new Specimen();
int[] path = new int[_matrix.Length + 1];
int cost = 0;
HashSet <int> vortexes = new HashSet <int>();
path[0] = 0;
for (int i = 1; i < _matrix.Length; i++)
{
vortexes.Add(i);
}
int prevVortex = 0;
for (int i = 1; i < _matrix.Length; i++)
{
int bestS = int.MaxValue;
int bestV = 0;
foreach (int x in vortexes)
{
if (_matrix[prevVortex][x] < bestS)
{
bestS = _matrix[prevVortex][x];
bestV = x;
}
}
path[i] = bestV;
cost += _matrix[prevVortex][bestV];
prevVortex = bestV;
vortexes.Remove(bestV);
}
path[path.Length - 1] = 0;
cost += _matrix[prevVortex][0];
toReturn.Path = path;
toReturn.Cost = cost;
return(toReturn);
}
private Specimen[] Crossing(Specimen[] population)
{
Specimen[] newPopulation = new Specimen[_populationSize];
int survivors = 0; //ilość osobników która przejdzie do następnego pokolenia, poprzez brak krzyżowania
int[] ruller = new int[_populationSize]; //tablica przechowywująca szanse na wylosowanie
//Uzupełnienie tej tablicy w sposób, który faworyzuje lepsze osobniki
ruller[0] = population[population.Length - 1].Cost - population[0].Cost;
for (int i = 1; i < _populationSize; i++)
{
ruller[i] = ruller[i - 1] + population[population.Length - 1].Cost - population[i].Cost;
}
for (int i = 0; i < population.Length / 2; i++)
{
//losowanie pierwszego osobnika do krzyżowania i wyszukanie jego indeksu
int random1, random2;
random1 = _random.Next(ruller[ruller.Length - 1]);
if (random1 < ruller[0])
{
random1 = 0;
}
else
{
for (int j = 1; j < ruller.Length; j++)
{
if (random1 < ruller[j])
{
random1 = j;
break;
}
}
}
//losowanie drugiego innego osobnika
random2 = _random.Next(ruller[ruller.Length - 1]);
if (random2 < ruller[0])
{
random2 = 0;
}
else
{
for (int j = 1; j < ruller.Length; j++)
{
if (random2 < ruller[j])
{
random2 = j;
break;
}
}
}
if (random2 == random1)
{
random2++;
}
if (random2 >= population.Length)
{
random2 = 0;
}
if (_random.Next(101) > _crossingFactor) //losowanie tego czy osobniki się skrzyżują lub nie
{
switch (_crossingMethod)
{
case 1:
PMXCrossing(population[random1].Path, population[random2].Path, out newPopulation[2 * i],
out newPopulation[2 * i + 1]);
break;
case 2:
OXCrossing(population[random1].Path, population[random2].Path, out newPopulation[2 * i], out newPopulation[2 * i + 1]);
break;
}
}
else
{
survivors += 2;
}
}
for (int i = 0; i < newPopulation.Length; i++)
{
if (newPopulation[i] == null)
{
newPopulation[i] = population[_random.Next(population.Length)];
}
}
return(newPopulation);
}
private void OXCrossing(int[] parrent1, int[] parrent2, out Specimen offspring1Out, out Specimen offspring2Out)
{
int[] offspring1 = new int[parrent1.Length];
int[] offspring2 = new int[parrent1.Length];
ArrayList genom1 = new ArrayList();
ArrayList genom2 = new ArrayList();
int a = 0, b = 0;
//losowanie 2 różnych punktów do krzyżownia
do
{
a = _random.Next(parrent1.Length - 2) + 1;
b = _random.Next(parrent1.Length - 2) + 1;
if (a > b)
{
int t = a;
a = b;
b = t;
}
} while (a == b);
for (int i = a; i <= b; i++)
{
offspring1[i] = parrent2[i];
offspring2[i] = parrent1[i];
genom1.Add(parrent2[i]);
genom2.Add(parrent1[i]);
}
//iterrators
int offspringit = b + 1, parrentit = b + 1;
//towrzenie potomka 1
while (offspringit < _size)
{
if (!genom1.Contains(parrent1[parrentit]))
{
offspring1[offspringit] = parrent1[parrentit];
offspringit++;
parrentit++;
}
else
{
parrentit++;
}
if (parrentit == _size)
{
parrentit = 1;
}
}
if (parrentit == _size)
{
parrentit = 1;
}
offspringit = 1;
while (offspringit < a)
{
if (!genom1.Contains(parrent1[parrentit]))
{
offspring1[offspringit] = parrent1[parrentit];
offspringit++;
parrentit++;
}
else
{
parrentit++;
}
}
//tworzenie potomka 2
offspringit = b + 1;
parrentit = b + 1;
while (offspringit < _size)
{
if (!genom2.Contains(parrent2[parrentit]))
{
offspring2[offspringit] = parrent2[parrentit];
offspringit++;
parrentit++;
}
else
{
parrentit++;
}
if (parrentit == _size)
{
parrentit = 1;
}
}
if (parrentit == _size)
{
parrentit = 1;
}
offspringit = 1;
while (offspringit < a)
{
if (!genom2.Contains(parrent2[parrentit]))
{
offspring2[offspringit] = parrent2[parrentit];
offspringit++;
parrentit++;
}
else
{
parrentit++;
}
}
offspring1Out = new Specimen();
offspring2Out = new Specimen();
offspring1Out.Path = offspring1;
offspring2Out.Path = offspring2;
offspring1Out.Cost = rateRute(offspring1);
offspring2Out.Cost = rateRute(offspring2);
}
private void PMXCrossing(int[] parrent1, int[] parrent2, out Specimen offspring1Out, out Specimen offspring2Out)
{
int[] offspring1 = new int[parrent1.Length];
int[] offspring2 = new int[parrent1.Length];
ArrayList genom1 = new ArrayList();
ArrayList genom2 = new ArrayList();
int a = 0, b = 0;
//losowanie 2 różnych punktów do krzyżownia
do
{
a = _random.Next(parrent1.Length - 2) + 1;
b = _random.Next(parrent1.Length - 2) + 1;
if (a > b)
{
int t = a;
a = b;
b = t;
}
} while (a == b);
for (int i = a; i <= b; i++)
{
offspring1[i] = parrent2[i];
offspring2[i] = parrent1[i];
genom2.Add(parrent1[i]);
genom1.Add(parrent2[i]);
}
for (int i = 0; i < a; i++)
{
if (!genom1.Contains(parrent1[i]))
{
offspring1[i] = parrent1[i];
}
if (!genom2.Contains(parrent2[i]))
{
offspring2[i] = parrent2[i];
}
}
for (int i = b + 1; i < parrent1.Length; i++)
{
if (!genom1.Contains(parrent1[i]))
{
offspring1[i] = parrent1[i];
}
if (!genom2.Contains(parrent2[i]))
{
offspring2[i] = parrent2[i];
}
}
for (int i = 1; i < a; i++)
{
if (offspring1[i] == 0)
{
offspring1[i] = (int)genom2[genom1.IndexOf(parrent1[i])];
while (genom1.Contains(offspring1[i]))
{
offspring1[i] = (int)genom2[genom1.IndexOf(offspring1[i])];
}
}
if (offspring2[i] == 0)
{
offspring2[i] = (int)genom1[genom2.IndexOf(parrent2[i])];
while (genom2.Contains(offspring2[i]))
{
offspring2[i] = (int)genom1[genom2.IndexOf(offspring2[i])];
}
}
}
for (int i = b + 1; i < parrent1.Length - 1; i++)
{
if (offspring1[i] == 0)
{
offspring1[i] = (int)genom2[genom1.IndexOf(parrent1[i])];
while (genom1.Contains(offspring1[i]))
{
offspring1[i] = (int)genom2[genom1.IndexOf(offspring1[i])];
}
}
if (offspring2[i] == 0)
{
offspring2[i] = (int)genom1[genom2.IndexOf(parrent2[i])];
while (genom2.Contains(offspring2[i]))
{
offspring2[i] = (int)genom1[genom2.IndexOf(offspring2[i])];
}
}
}
offspring1Out = new Specimen();
offspring2Out = new Specimen();
offspring1Out.Path = offspring1;
offspring2Out.Path = offspring2;
offspring1Out.Cost = rateRute(offspring1);
offspring2Out.Cost = rateRute(offspring2);
}
public static void run()
{
ArrayList scores = new ArrayList();
for (int file = 0; file < 3; file++)
{
scores = new ArrayList();
Specimen fileBest = new Specimen();
fileBest.Cost = int.MaxValue;
MatrixReader mr = new MatrixReader();
mr.load(Variables.Paths[file]);
Console.WriteLine(mr.name);
for (int crossParam = 1; crossParam <= 2; crossParam++)
{
for (int mutatuinParam = 1; mutatuinParam <= 2; mutatuinParam++)
{
for (int PopulationSizeIndex = 0; PopulationSizeIndex < 3; PopulationSizeIndex++)
{
for (int ttl = 40; ttl <= 120; ttl += 40)
{
int cost = 0;
for (int i = 0; i < 3; i++)
{
TSPGenetic tsp = new TSPGenetic(mr.Matrix, ttl,
Variables.PopulationSize[PopulationSizeIndex], 0.01f, 0.8f, mutatuinParam,
crossParam);
tsp.run();
cost += tsp.GetBestSpecimen().Cost;
if (tsp.GetBestSpecimen().Cost < fileBest.Cost)
{
fileBest = tsp.GetBestSpecimen();
}
}
scores.Add(new Score(mutatuinParam, crossParam, ttl, cost / 3, 0.01f, 0.8f, Variables.PopulationSize[PopulationSizeIndex]));
Console.WriteLine(((Score)scores[scores.Count - 1]).CrossingType + " " + ((Score)scores[scores.Count - 1]).MutationType + " " + ((Score)scores[scores.Count - 1]).PopulationSize);
}
}
}
}
using (StreamWriter streamWriter = File.CreateText(@"D:\tsp3\wyniki\pkt2" + mr.name + ".txt"))
{
var writer = new CsvWriter(streamWriter);
writer.Configuration.RegisterClassMap <ScoreMap>();
writer.WriteRecords(scores);
}
//wybieranie najlepszego trymu mutacji i krzyżowania
Score best = new Score(1, 1, 1, int.MaxValue, 0f, 0f, 0);
foreach (Score score in scores)
{
if (score.Cost < best.Cost)
{
best = (Score)score;
}
}
int bestmutator;
int bestcrosser;
if (best.CrossingType.Equals("PMX"))
{
bestcrosser = 1;
}
else
{
bestcrosser = 2;
}
if (best.MutationType.Equals("Invert"))
{
bestmutator = 1;
}
else
{
bestmutator = 2;
}
scores = new ArrayList();
float[] mutators = { 0.01f, 0.05f, 0.1f };
float[] crossators = { 0.5f, 0.7f, 0.9f };
for (int mutator = 0; mutator < 3; mutator++)
{
int cost = 0;
for (int i = 0; i < 3; i++)
{
TSPGenetic tsp = new TSPGenetic(mr.Matrix, 60, best.PopulationSize, mutators[mutator], 0.8f, bestmutator, bestcrosser);
tsp.run();
cost += tsp.GetBestSpecimen().Cost;
if (tsp.GetBestSpecimen().Cost < fileBest.Cost)
{
fileBest = tsp.GetBestSpecimen();
}
}
scores.Add(new Score(bestmutator, bestcrosser, 60, cost / 3, mutators[mutator], 0.8f,
best.PopulationSize));
Console.WriteLine("Mutation " + mutator);
}
using (StreamWriter streamWriter = File.CreateText(@"D:\tsp3\wyniki\pkt3" + mr.name + ".txt"))
{
var writer = new CsvWriter(streamWriter);
writer.Configuration.RegisterClassMap <ScoreMap>();
writer.WriteRecords(scores);
}
scores = new ArrayList();
for (int crossator = 0; crossator < 3; crossator++)
{
int cost = 0;
for (int i = 0; i < 3; i++)
{
TSPGenetic tsp = new TSPGenetic(mr.Matrix, 60, best.PopulationSize, 0.01f, crossators[crossator], bestmutator, bestcrosser);
tsp.run();
cost += tsp.GetBestSpecimen().Cost;
if (tsp.GetBestSpecimen().Cost < fileBest.Cost)
{
fileBest = tsp.GetBestSpecimen();
}
}
scores.Add(new Score(bestmutator, bestcrosser, 60, cost / 3, 0.01f, crossators[crossator],
best.PopulationSize));
Console.WriteLine("Crossing " + crossator);
}
using (StreamWriter streamWriter = File.CreateText(@"D:\tsp3\wyniki\pkt4" + mr.name + ".txt"))
{
var writer = new CsvWriter(streamWriter);
writer.Configuration.RegisterClassMap <ScoreMap>();
writer.WriteRecords(scores);
}
using (StreamWriter streamWriter = File.CreateText(@"D:\tsp3\wyniki\pkt5" + mr.name + ".txt"))
{
streamWriter.WriteLine("Cost: " + fileBest.Cost + "\n");
foreach (var i in fileBest.Path)
{
streamWriter.Write(i + "=>");
}
streamWriter.Write("\b");
streamWriter.Write("\b");
}
}
}