public TSPOrganism crossover(TSPOrganism parent)
{
int start = 0;
int end = 0;
int diff = 0;
// Define a range
while (true)
{
// Obtain two indices to swap
start = rnd.Next(0, NUM_CITIES);
end = rnd.Next(0, NUM_CITIES);
if (start == end)
{
continue;
}
// How much did we chop off?
diff = Math.Abs(start - end);
double percentage = (double)diff / route.Count;
if (percentage <= MAX_CROSSOVER_PERCENTAGE)
{
break;
}
}
// Force the start to be the lower number, and end to be the higher
// this is mainly just for sanity's sake
if (start > end)
{
int temp = start;
start = end;
end = temp;
}
// Now for the fun bit!
ArrayList childRoute = new ArrayList();
childRoute.AddRange(route.GetRange(start, diff)); // Add the subsequence to the child's list
// Now, loop through the second parent, starting after 'end'
// If the element is not contained in the child route, we add it (to the end)
for (int i = end + 1; i < NUM_CITIES; i++)
{
City currentCity = parent.getRoute()[i] as City;
if (childRoute.Contains(currentCity))
{
continue;
}
childRoute.Add(currentCity);
}
// Then we have to add the cities before 'end' that we may have missed
ArrayList prefix = new ArrayList();
for (int i = 0; i < end + 1; i++)
{
City currentCity = parent.getRoute()[i] as City;
if (childRoute.Contains(currentCity))
{
continue;
}
//childRoute.Insert(0, currentCity);
prefix.Add(currentCity);
}
// Add the pre-pended portion
childRoute.InsertRange(0, prefix);
return(new TSPOrganism(childRoute));
}
public string[] fancySolveProblem()
{
string[] results = new string[3];
/* Algorithm Parameters */
int populationSize = 100;
long maxGenerations = 5000;
long stagnationCutoff = 300;
bool shouldTimeout = true;
// These are some dealios that you can customize
// Apparently these are pretty statndard values
// But really I know nothing about the theory behind this
double percentMated = 0.6;
double percentMutated = 0.3;
double percentRandom = 0.1;
// Percent of the initial population that should be a clone of the greedy solution
double greedyProportion = 0.25;
// Set up our Organism class
TSPOrganism.Cities = Cities;
TSPOrganism.NUM_CITIES = Cities.Length;
TSPOrganism.isOnlyValidRoutes = true; // Toggle if infinity cost routes are allowed
//TSPOrganism.setSeed(_seed); // Used for debugging, mainly
// Create the initial list
MotherNature motherNature = new MotherNature(populationSize, percentMated, percentMutated, percentRandom);
Stopwatch timer = new Stopwatch();
timer.Start();
// Start by Obtaining the Greedy Solution
greedySolveProblem();
// Convert the results to a TSPOrganism
TSPOrganism greedySeed = new TSPOrganism(bssf.Route);
List <TSPOrganism> population = motherNature.generatePopulationFromOrganism(greedySeed, greedyProportion);
// Define some counters or whatever
long numGenerations = 0;
long gensSinceLastBest = 0;
int numNewBest = 0;
//TSPOrganism bestSoFar = TSPOrganism.newOrganism(); // Start with a random best
TSPOrganism bestSoFar = greedySeed;
// Here's acutal interesting code
while (true)
{
if (timer.ElapsedMilliseconds / 1000 > TIME_LIMIT)
{
break;
}
numGenerations++;
population = population.OrderBy(a => a.getFitness()).ToList();
List <TSPOrganism> fittestParents = population.GetRange(0, populationSize / 2);
// New best?
if (fittestParents[0].getFitness() < bestSoFar.getFitness())
{
bestSoFar = fittestParents[0];
gensSinceLastBest = 0;
numNewBest++;
// Update our ratios
// Start to favor mutation more and more
//percentMutated += ((1 - percentMutated) / ((0.02449 * Cities.Length) + 4.122));
percentMutated += ((1 - percentMutated) / (1 + Math.Log(Cities.Length)));
// Then update our percent mated
percentMated = 1 - (percentMutated + percentRandom);
motherNature.setPercentMutated(percentMutated);
motherNature.setPercentMated(percentMated);
}
else
{
gensSinceLastBest++;
}
List <TSPOrganism> offspring = motherNature.createOffspring(fittestParents);
// Find the fittest Offspring
offspring = offspring.OrderBy(a => a.getFitness()).ToList();
List <TSPOrganism> fittestOffspring = offspring.GetRange(0, populationSize / 2);
// Merge best parents and offspring to create the new population
population = new List <TSPOrganism>();
population.AddRange(fittestParents);
population.AddRange(fittestOffspring);
if (shouldTimeout)
{
continue;
}
// Otherwise, test our break conditions
if (gensSinceLastBest >= stagnationCutoff)
{
break;
}
if (numGenerations >= maxGenerations)
{
break;
}
}
timer.Stop();
// Convert our solution
bssf = new TSPSolution(bestSoFar.getRoute());
results[COST] = Convert.ToString(bestSoFar.getFitness());
results[TIME] = timer.Elapsed.ToString();
results[COUNT] = Convert.ToString(numNewBest);
return(results);
}
