private IList <TaskPair> FindParents()
{
ICollection <TaskPair> parentSet = new SortedSet <TaskPair>
{
new TaskPair(IndividualTask.X_start, IndividualTask.count_fx(IndividualTask.X_start))
};
for (int i = 0; i < parentsQuantity - 1; i++)
{
TaskPair newTaskPair = new TaskPair();
int[] x_new = new int[n];
Array.Copy(IndividualTask.X_start, x_new, n);
for (int j = 0; j < n; j++)
{
int boolval = rand.Next(2);
if (boolval == 1)
{
x_new[j] += rand.Next(
Math.Max(-5, (IndividualTask.Constraints[j][0] - IndividualTask.X_start[j])),
Math.Min(6, (IndividualTask.Constraints[j][1] - IndividualTask.X_start[j])));
}
}
newTaskPair.X = x_new;
newTaskPair.Fx = IndividualTask.count_fx(x_new);
parentSet.Add(newTaskPair);
}
parentsQuantity = parentsQuantity / 3;
if (parentsQuantity % 2 != 0)
{
parentsQuantity++;
}
return(parentSet.Take(parentsQuantity).ToList());
}
public TaskPair ExecuteAlgorithm()
{
IList <TaskPair> parents = FindParents();
long record = parents.First().Fx;
int lasting = 0;
TaskPair taskPair = get_children(parents, record, lasting);
return(taskPair);
}
private TaskPair ExecuteGeneticAndSaveResults(Task task)
{
Genetic genetic = new Genetic(task, 100);
timeCounter = Stopwatch.StartNew();
TaskPair pair1 = genetic.ExecuteAlgorithm();
timeCounter.Stop();
geneticTime.Add(timeCounter.ElapsedTicks);
return(pair1);
}
private TaskPair ExecuteAnnealingAndSaveResults(Task task)
{
SimulatedAnnealing simulatedAnnealing = new SimulatedAnnealing(task, 200, 10);
timeCounter = Stopwatch.StartNew();
TaskPair pair2 = simulatedAnnealing.executeAlgorithm();
timeCounter.Stop();
annealingTime.Add(timeCounter.ElapsedTicks);
return(pair2);
}
private TaskPair generateChild(TaskPair parent1, TaskPair parent2) //генеруємо нащадка
{
TaskPair child = new TaskPair();
child.X = new int[this.n];
for (int i = 0; i < n; i++)
{
double alpha = Convert.ToDouble(rand.Next(-25, 125)) / 100;
child.X[i] = Convert.ToInt32(parent1.X[i] + alpha * (parent2.X[i] - parent1.X[i]));
}
child.Fx = IndividualTask.count_fx(child.X);
return(child);
}
private bool balanceIsReached(IList <TaskPair> points, TaskPair currentPoint, float epsilon)
{
IList <double> Ratio = new List <double>();
foreach (TaskPair point in points)
{
if (point != currentPoint)
{
Ratio.Add(Convert.ToDouble(Math.Abs(point.Fx - currentPoint.Fx)) / point.Fx);
}
}
return((Ratio.Count != 0)?(Ratio.Min() <= epsilon):false);
}
private bool childIsAdmissible(TaskPair child)
{
bool admissible = true;
for (int i = 0; i < n; i++)
{
if (child.X[i] < IndividualTask.Constraints[i][0] ||
child.X[i] > IndividualTask.Constraints[i][1])
{
admissible = false;
break;
}
}
return(admissible);
}
public CounterResult CountTime(int i, int productsQuantity, int leftSum, int rightSum, int leftAmount, int rightAmount)
{
List <TaskPair[]> Fxs = new List <TaskPair[]>();
CounterResult result = new CounterResult();
//int i = 1000; //user-defined number of Individual Tasks to generate
int geneticBetter = 0;
int annealBetter = 0;
int wolfBetter = 0;
//int productsQuantity = 3; //may be randomized or user-defined
while (i != 0)
{
Order generatedOrder = RandomHandler.GetRandomizedOrder(productsQuantity, leftSum, rightSum, leftAmount, rightAmount);
Task task = ConvertOrderToTask(generatedOrder);
TaskPair pair1 = ExecuteGeneticAndSaveResults(task);
TaskPair pair2 = ExecuteAnnealingAndSaveResults(task);
TaskPair pair3 = ExecuteFrankWolfAndSaveResults(generatedOrder, task);
TaskPair[] pairs = new TaskPair[3] {
pair1, pair2, pair3
};
//counting situations when one algorithm gives results better than other
//(you can remove it or use to compare results)
if (pair1.Fx != 0 || pair2.Fx != 0 || pair3.Fx != 0)
{
if (pair1.Fx <= pair2.Fx && pair1.Fx <= pair3.Fx)
{
geneticBetter++;
}
if (pair2.Fx <= pair1.Fx && pair2.Fx <= pair3.Fx)
{
annealBetter++;
}
if (pair3.Fx <= pair1.Fx && pair3.Fx <= pair2.Fx)
{
wolfBetter++;
}
}
Fxs.Add(pairs);
i--;
}
result.AnnealAvg = annealingTime.Average();
result.GeneticAvg = geneticTime.Average();
result.WolfAvg = wolfTime.Average();
result.GeneticBetter = geneticBetter;
result.AnnealBetter = annealBetter;
result.FrankBetter = wolfBetter;
return(result);
}
public TaskPair executeAlgorithm()
{
int temperature = startTemperature;
int h = 0;
float epsilon = 0.01F;
int[] x_current = new int[n];
Array.Copy(IndividualTask.X_start, x_current, n);
TaskPair bestSolution = new TaskPair(x_current, IndividualTask.count_fx(x_current));
TaskPair currentSolution = bestSolution;
while (h <= H && bestSolution.Fx != 0 && temperature != 0)
{
IList <TaskPair> z = new List <TaskPair>();
while (bestSolution.Fx != 0 && !balanceIsReached(z, currentSolution, epsilon))
{
z.Add(currentSolution);
TaskPair Y = generateNewPoint(x_current);
if (Y.Fx == 0)
{
bestSolution = Y;
break;
}
long delta = Y.Fx - currentSolution.Fx;
double p = Math.Min(1, Math.Pow(Math.E, (-delta / temperature)));
int minKsi = 75 + (75 * (temperature - startTemperature) / startTemperature);
int maxKsi = 100 * temperature / startTemperature;
double ksi = Convert.ToDouble(rand.Next(minKsi, maxKsi + 1)) / 100;
if (p > ksi || delta < 0)
{
currentSolution = Y;
if (Y.Fx < bestSolution.Fx)
{
bestSolution = Y;
}
}
}
h++;
double alpha = Convert.ToDouble(rand.Next(50, 100)) / 100;
temperature = Convert.ToInt32(temperature * alpha);
}
return(bestSolution);
}
private TaskPair get_children(IList <TaskPair> parents, long record, int lasting)
{
if (parents.Count == 1)
{
return(parents.ElementAt(0));
}
if (parents.Count % 2 != 0)
{
parents.RemoveAt(parents.Count - 1);
}
ICollection <TaskPair> iterationNodes = new SortedSet <TaskPair>();
foreach (TaskPair parent in parents)
{
iterationNodes.Add(parent);
}
TaskPair.Shuffle(parents);
for (int k = 0; k < parents.Count; k += 2)
{
int quantity = rand.Next(1, 5);
while (quantity != 0)
{
TaskPair child = generateChild(parents[k], parents[k + 1]);
ChildMutation(child);
if (childIsAdmissible(child))
{
child.Fx = IndividualTask.count_fx(child.X);
iterationNodes.Add(child);
quantity--;
}
}
}
long newRecord = iterationNodes.First().Fx;
if ((newRecord == record && lasting == 99) || newRecord == 0)
{
return(iterationNodes.First());
}
else
{
IList <TaskPair> iterationNodesList = iterationNodes
.Take(3 * iterationNodes.Count / 4)
.ToList();
IList <TaskPair> new_nodes = iterationNodes
.Take(parentsQuantity / 2)
.ToList();
foreach (TaskPair addedNode in new_nodes)
{
iterationNodesList.Remove(addedNode);
}
TaskPair.Shuffle(iterationNodesList);
for (int i = 0; i < Math.Min(parentsQuantity / 2, iterationNodesList.Count); i++)
{
new_nodes.Add(iterationNodesList.ElementAt(i));
}
if (new_nodes.Count % 2 != 0)
{
new_nodes.RemoveAt(new_nodes.Count - 1);
}
if (newRecord == record)
{
return(get_children(new_nodes, newRecord, lasting + 1));
}
else
{
return(get_children(new_nodes, newRecord, 0));
}
}
}
private void ChildMutation(TaskPair child)
{
int IdxToMutate = rand.Next(n);
child.X[IdxToMutate]--;
}