public int run(GeneticRunParameters parameters)
{
if (!(parameters is RandomRestartRunParameters))
throw new WrongParameterTypeException("Random Restart Genetics can only be run using Random Restart Parameters");
RandomRestartRunParameters rrParameters = (RandomRestartRunParameters)parameters;
int bestFitness = 0;
List<GeneStringFitnessPair> bestResults = new List<GeneStringFitnessPair>();
do
{
subGenetics = rrParameters.algorithm.newInstance(initial, rrParameters.algorithmInstanceParameters);
int maxfitness = runUntilRepeating(rrParameters.algorithmRunParameters, rrParameters.numRepeatsBeforeRestart, rrParameters.verbose);
GeneStringFitnessPair bestOfGeneration = subGenetics.getBestPair();
bestResults.Add(bestOfGeneration);
if (maxfitness > bestFitness) bestFitness = maxfitness;
if (rrParameters.verbose)
{
Console.WriteLine();
Console.Write("Local Max at " + maxfitness);
}
goalValue = (1000 - rrParameters.promilleGoalDecreasePerRound) * goalValue / 1000;
if (rrParameters.verbose) Console.WriteLine(", restarting with goal at " + goalValue + " best = " + bestFitness);
} while (bestFitness < goalValue);
return runChampionRound(rrParameters, bestResults);
}
public RandomRestartRunParameters(GeneticAlgorithm algorithm, GeneticInstanceParameters algorithmInstanceParameters,
GeneticRunParameters algorithmRunParameters, int promilleGoalDecreasePerRound, int numRepeatsBeforeRestart, bool verbose)
{
this.algorithm = algorithm;
this.algorithmInstanceParameters = algorithmInstanceParameters;
this.algorithmRunParameters = algorithmRunParameters;
this.promilleGoalDecreasePerRound = promilleGoalDecreasePerRound;
this.numRepeatsBeforeRestart = numRepeatsBeforeRestart;
this.verbose = verbose;
}
public FuzzyGeneticsRunParameters(GeneticAlgorithm algorithm, GeneticInstanceParameters algorithmInstanceParameters,
GeneticRunParameters algorithmRunParameters, int numberToAverageScoreOver, int waitSinceLastImprovement, bool verbose)
{
this.algorithm = algorithm;
this.algorithmInstanceParameters = algorithmInstanceParameters;
this.algorithmRunParameters = algorithmRunParameters;
this.numberToAverageScoreOver = numberToAverageScoreOver;
this.waitSinceLastImprovement = waitSinceLastImprovement;
this.verbose = verbose;
}
public int run(GeneticRunParameters parameters)
{
if (!(parameters is FuzzyGeneticsRunParameters))
throw new WrongParameterTypeException("Fuzzy Genetics only accepts Fuzzy Genetics Run Parameters");
FuzzyGeneticsRunParameters fgParameters = (FuzzyGeneticsRunParameters)parameters;
subGenetics = fgParameters.algorithm.newInstance(initial, fgParameters.algorithmInstanceParameters);
int bestFitness = 0;
int periodSinceLastImprovement = 0;
LinkedList<int> lastFitnesses = new LinkedList<int>();
for (int n = 0; n < fgParameters.numberToAverageScoreOver; n++) { lastFitnesses.AddFirst(0); }
while (periodSinceLastImprovement < fgParameters.waitSinceLastImprovement)
{
int fitness = subGenetics.run(fgParameters.algorithmRunParameters);
lastFitnesses.RemoveFirst();
lastFitnesses.AddLast(fitness);
int summedFitness = lastFitnesses.Sum();
periodSinceLastImprovement++;
if (bestFitness < summedFitness)
{
bestFitness = summedFitness;
periodSinceLastImprovement = 0;
result = subGenetics.getAll();
resultPairs = subGenetics.getAllPairs();
}
if (fgParameters.verbose) Console.WriteLine("best of round: " + fitness + ". average = " + summedFitness / fgParameters.numberToAverageScoreOver +
" (" + periodSinceLastImprovement + ")");
}
return lastFitnesses.Last.Value;
}
// remainder of 1 - mutation - crossover = copy chance
// returns max fitness this round
public int run(GeneticRunParameters parameters)
{
if (!(parameters is BasicGeneticRunParameters)) throw new WrongParameterTypeException("Basic Genetics only take Basic Genetics Run Parameters");
BasicGeneticRunParameters bParameters = (BasicGeneticRunParameters)parameters;
if (bParameters.mutationChance + bParameters.crossoverChance > 1.0d)
throw new OutOfBoundsException("mutation chance + crossover chance is more than 100%");
double crossoverChance = bParameters.crossoverChance + bParameters.mutationChance; // < mutation = mutation, else < crossover = crossover, else copy
GeneString[] newGeneration = new GeneString[population.Count];
for (int n = 0; n < bParameters.keepPerRound; n++)
{
newGeneration[n] = population[n].genestring.copy();
}
for (int n = bParameters.keepPerRound; n < population.Count; n++)
{
Double rnd = SharedRandom.NextDouble();
if (rnd < bParameters.mutationChance) newGeneration[n] = mutateFrom(bParameters.keepPerRound, bParameters.mutationSeverity);
else if (rnd < crossoverChance) newGeneration[n] = crossoverFrom(bParameters.keepPerRound, bParameters.keepLength);
else newGeneration[n] = copyFrom(bParameters.keepPerRound);
}
population = evaluateAll(newGeneration);
population.Sort();
return population[0].fitness;
}
private int runUntilRepeating(GeneticRunParameters parameters, int numRepeatsBeforeRestart, Boolean verbose)
{
int numAtSameMax = 0;
int currentMax = 0;
int maxFitness = 0;
while (numAtSameMax != numRepeatsBeforeRestart)
{
maxFitness = subGenetics.run(parameters);
if (maxFitness == currentMax) numAtSameMax++;
else
{
numAtSameMax = 0;
if (verbose) Console.Write(".");
}
currentMax = maxFitness;
}
return maxFitness;
}
