public override IEntity Decode(MultiKey islandKey, Dictionary <MultiKey, IEntity> entities)
{
double totalDistance = 0.0;
var entity = entities[Key] as TSPEntity;
var chromosome = entity.Chromosomes[Key[0]] as PermutationChromosome;
var firstPoint = TSPPoints[chromosome.Genes[0]];
var previousPoint = firstPoint;
for (int i = 1; i < chromosome.Genes.Length; i++)
{
var point = TSPPoints[chromosome.Genes[i]];
totalDistance += GetDistance(previousPoint, point);
entity.Phenotype.Add(previousPoint);
previousPoint = point;
}
totalDistance += GetDistance(previousPoint, firstPoint);
entity.Phenotype.Add(previousPoint);
MultiObjectiveFitness fitness = new MultiObjectiveFitness(new double[] { -1 * totalDistance });
entity.SetFitness(fitness);
entity.TotalDistance = totalDistance;
EntityCount++;
return(entity);
}
public override IEntity Decode(MultiKey islandKey, Dictionary <MultiKey, IEntity> entities)
{
EntityCount++;
SortedSubsetChromosomeValidator.EntityCount = EntityCount;
var entity = entities[Key] as VehicleSchedulingEntity;
var chromosome = entity.Chromosomes[Key[0]] as SortedSubsetChromosome;
entity.VehiclesCount = chromosome.Sections.Length;
for (int s = 0; s < chromosome.Sections.Length; s++)
{
//int DailyFirstTime = InitData.Trips[chromosome.Sections[s][0]].ArrivalTime;
//int length = chromosome.Sections[s].Length;
//int DailyLastTime = InitData.Trips[chromosome.Sections[s][length-1]].DepartureTime;
//int? AMLastTime = null;
//int? PMFirstTime = null;
for (int p = 0; p < chromosome.Sections[s].Length - 1; p++)
{
if (ConflictDetector.ConflictDetected(chromosome.Sections[s][p], chromosome.Sections[s][p + 1]))
{
throw new ApplicationException("Shit happend");
//return null; //hard conflict
}
var trip1 = InitData.Trips[chromosome.Sections[s][p]];
var trip2 = InitData.Trips[chromosome.Sections[s][p + 1]];
double distance = InitData.GetDistance(trip1.LastStopId, trip2.FirstStopId);
entity.TotalDeadMileage += distance;
//if ((trip1.DepartureTime <= 720) && (!AMLastTime.HasValue || trip1.DepartureTime > AMLastTime)) AMLastTime = trip1.DepartureTime;
//if ((trip1.ArrivalTime >= 720) && (!PMFirstTime.HasValue || trip1.ArrivalTime < PMFirstTime)) PMFirstTime = trip1.ArrivalTime;
}
//var lastTrip = InitData.Trips[chromosome.Sections[s][length-1]];
//if ((lastTrip.DepartureTime <= 720) && (!AMLastTime.HasValue || lastTrip.DepartureTime > AMLastTime)) AMLastTime = lastTrip.DepartureTime;
//if ((lastTrip.ArrivalTime >= 720) && (!PMFirstTime.HasValue || lastTrip.ArrivalTime < PMFirstTime)) PMFirstTime = lastTrip.ArrivalTime;
//entity.TotalActiveTime += DailyLastTime - DailyFirstTime;
//if (AMLastTime.HasValue) entity.TotalActiveTime -= (720 - AMLastTime.Value);
//if (PMFirstTime.HasValue) entity.TotalActiveTime -= (PMFirstTime.Value - 720);
}
var fitness0 = 1 / (1 + entity.TotalDeadMileage);
var fitness1 = 1 / (1 + (double)entity.VehiclesCount);
MultiObjectiveFitness fitness = new MultiObjectiveFitness(new double[] { fitness0, fitness1 });
//fitness.Value[2] = 1 / (1 + (double) entity.TotalActiveTime);
//GetAverageLengthOfLongSections(chromosome.Sections);
entity.SetFitness(fitness);
return(entity);
}
static void Main(string[] args)
{
var tspData = LoadCsv("Berlin52.csv");
var initData = new TSPInitData(tspData);
var optimizer = Optimizer.Create();
optimizer.Settings.WithSeed(Environment.TickCount);
optimizer.Settings.AddSubProblem("TSP", new TravelingSalesman(tspData.Count));
optimizer.Settings.WithEntityType <TSPEntity>().WithEvaluation <TSPEvaluation>();
optimizer.SetParameter("TestParameter", 42.0);
var fitnessLimit = new MultiObjectiveFitness(new double[] { -7545 });
optimizer.Settings.StopWhen().FitnessLimitExceeded(fitnessLimit)
.Or().TimeoutElapsed(300000);
optimizer.SetParameter(Pea.Core.Island.ParameterNames.IslandsCount, 10);
Stopwatch sw = Stopwatch.StartNew();
var result = optimizer.Run(initData);
//var result = AsyncUtil.RunSync(() => system.Start(initData));
foreach (var reason in result.StopReasons)
{
Console.WriteLine(reason);
}
sw.Stop();
var elapsed = sw.ElapsedMilliseconds;
var entities = TSPEvaluation.EntityCount;
var speed = entities / elapsed;
Console.WriteLine($"Elapsed: {elapsed} Entities: {entities} ({speed} ent./ms)");
Console.ReadLine();
}
public void BuildArchive(IEvolutionState state, IList <Individual> oldInds, IList <Individual> newInds, int archiveSize)
{
// step 1: load the archive with the pareto-nondominated front
var archive = new List <Individual>();
var nonFront = new List <Individual>();
MultiObjectiveFitness.PartitionIntoParetoFront(oldInds, archive, nonFront);
var currentArchiveSize = archive.Count;
// step 2: if the archive isn't full, load the remainder with the fittest individuals (using customFitnessMetric) that aren't in the archive yet
if (currentArchiveSize < archiveSize)
{
// BRS : The following uses Individual's IComparable implementation based on Fitness.
// The fitter individuals will be earlier
nonFront.SortByFitnessDescending();
var len = archiveSize - currentArchiveSize;
for (var i = 0; i < len; i++)
{
archive.Add(nonFront[i]);
currentArchiveSize++;
}
}
// step 3: if the archive is OVERFULL, iterate as follows:
// step 3a: remove the k-closest individual in the archive
//var evaluator = ((ISPEA2Evaluator)(state.Evaluator));
//var inds = archive.ToArray();
while (currentArchiveSize > archiveSize)
{
var closest = archive[0];
var closestIndex = 0;
var closestD = CalculateDistancesFromIndividual(closest, oldInds);
for (var i = 1; i < currentArchiveSize; i++)
{
var competitor = archive[i];
var competitorD = CalculateDistancesFromIndividual(competitor, oldInds);
for (var k = 0; k < oldInds.Count; k++)
{
if (closestD[i] > competitorD[i])
{
closest = competitor;
closestD = competitorD;
closestIndex = k;
break;
}
else if (closestD[i] < competitorD[i])
{
break;
}
}
}
// remove him destructively -- put the top guy in his place and remove the top guy. This is O(1)
archive[closestIndex] = archive[archive.Count - 1];
archive.RemoveAt(archive.Count - 1);
currentArchiveSize--;
}
// step 4: put clones of the archive in the new individuals
var arr = archive.ToArray();
for (var i = 0; i < archiveSize; i++)
{
newInds[newInds.Count - archiveSize + i] = (Individual)arr[i].Clone();
}
}
/// <summary>
/// Program entry point.
/// </summary>
/// <param name="app">Holds arguments and other info.</param>
public void Execute(IExampleInterface app)
{
IMLDataSet trainingData = GenerationUtil.GenerateSingleDataRange(
(x) => (3 * Math.Pow(x, 2) + (12 * x) + 4)
, 0, 100, 1);
EncogProgramContext context = new EncogProgramContext();
context.DefineVariable("x");
StandardExtensions.CreateNumericOperators(context);
PrgPopulation pop = new PrgPopulation(context, 1000);
MultiObjectiveFitness score = new MultiObjectiveFitness();
score.AddObjective(1.0, new TrainingSetScore(trainingData));
TrainEA genetic = new TrainEA(pop, score);
genetic.ValidationMode = true;
genetic.CODEC = new PrgCODEC();
genetic.AddOperation(0.5, new SubtreeCrossover());
genetic.AddOperation(0.25, new ConstMutation(context, 0.5, 1.0));
genetic.AddOperation(0.25, new SubtreeMutation(context, 4));
genetic.AddScoreAdjuster(new ComplexityAdjustedScore(10, 20, 10, 20.0));
genetic.Rules.AddRewriteRule(new RewriteConstants());
genetic.Rules.AddRewriteRule(new RewriteAlgebraic());
genetic.Speciation = new PrgSpeciation();
(new RampedHalfAndHalf(context, 1, 6)).Generate(new EncogRandom(), pop);
genetic.ShouldIgnoreExceptions = false;
EncogProgram best = null;
genetic.ThreadCount = 1;
try
{
for (int i = 0; i < 1000; i++)
{
genetic.Iteration();
best = (EncogProgram)genetic.BestGenome;
Console.Out.WriteLine(genetic.IterationNumber + ", Error: "
+ best.Score + ",Best Genome Size:" + best.Size
+ ",Species Count:" + pop.Species.Count + ",best: " + best.DumpAsCommonExpression());
}
//EncogUtility.evaluate(best, trainingData);
Console.Out.WriteLine("Final score:" + best.Score
+ ", effective score:" + best.AdjustedScore);
Console.Out.WriteLine(best.DumpAsCommonExpression());
//pop.dumpMembers(Integer.MAX_VALUE);
//pop.dumpMembers(10);
}
catch (Exception t)
{
Console.Out.WriteLine(t.ToString());
}
finally
{
genetic.FinishTraining();
EncogFramework.Instance.Shutdown();
}
}
public void MultiObjectiveFitnessWriteAndRead()
{
var rand = new MersenneTwisterFast(0);
var f = new MultiObjectiveFitness(2);
var f2 = new MultiObjectiveFitness(2); // We'll use this when we read the other one back in
f2.SetObjectives(null, new [] { 0.0, 0.0 }); // setting these to zero allows us to check if they change
// Default is to Maximize!
// These need to be set up manually here because we are not running 'Setup'.
// Every instance would thus normally share the same min and max values.
f.MaxObjective = new [] { 1.0, 1.0 };
f.MinObjective = new [] { 0.0, 0.0 };
f2.MaxObjective = new[] { 1.0, 1.0 };
f2.MinObjective = new[] { 0.0, 0.0 };
// Set two objective (fitness) values, worst and best respectively
f.SetObjectives(null, new [] { 0.0, 1.0 });
// Set up some random Trials just to check that they get transmitted
const int n = 10;
f.Trials = new List <double>(n);
for (var i = 0; i < n; i++)
{
f.Trials.Add(rand.NextDouble() * double.MaxValue); // in the half-open interval [0.0, double.MaxValue)
}
using (var ms = new MemoryStream())
{
var writer = new BinaryWriter(ms);
f.WriteFitness(null, writer); // Write
ms.Position = 0;
var reader = new BinaryReader(ms);
f2.ReadFitness(null, reader); // Read
// Compare
Assert.AreEqual(f.Value, f2.Value); // Value is same. This is the MAX objective value in the array
// Just for kicks, lets make sure BOTH objectives are equal to the original values
Assert.AreEqual(f.GetObjective(0), 0.0);
Assert.AreEqual(f.GetObjective(0), f2.GetObjective(0));
Assert.AreEqual(f.GetObjective(1), 1.0);
Assert.AreEqual(f.GetObjective(1), f2.GetObjective(1));
// And let's make sure our MAX and MIN are the same (these wouldn't normally change once established in 'Setup')
Assert.AreEqual(f.MinObjective[0], f2.MinObjective[0]);
Assert.AreEqual(f.MinObjective[1], f2.MinObjective[1]);
Assert.AreEqual(f.MaxObjective[0], f2.MaxObjective[0]);
Assert.AreEqual(f.MaxObjective[1], f2.MaxObjective[1]);
Assert.IsTrue(f.Value <= 1.0f && f2.Value <= 1.0f); // This is the MAX objective value in the array
Assert.IsFalse(f2.IsIdeal); // Fitness is not ideal by default (ideal must be defined in subclass)
Assert.AreEqual(f2.Trials.Count, f.Trials.Count); // Number of trials is the same
for (var i = 0; i < f.Trials.Count; i++)
{
Assert.AreEqual((double)f.Trials[i], (double)f2.Trials[i]); // Trial values are all the same
}
}
}
