internal Population RunGeneration(int currentGeneration, Population currentPopulation, FitnessFunction fitnessFunctionDelegate)
{
//create new empty populations for processing
var tempPopulation = currentPopulation.CreateEmptyCopy();
var processedPopulation = tempPopulation.CreateEmptyCopy();
tempPopulation.Solutions.AddRange(currentPopulation.Solutions);
//clear the current population, this keeps it simple as the solutions could
//easily get corrupted by genetic operators as from v 2.04 on, the genes are no longer cloned
currentPopulation.Solutions.Clear();
//invoke the operators
var enabledOperators = this.Operators.Where((o) => o.Enabled).ToList();
var operatorCount = enabledOperators.Count();
for (int op = 0; op < operatorCount; op++)
{
var enabled = true;
var genOp = enabledOperators [op] as IGeneticOperator;
if (genOp != null)
{
enabled = genOp.Enabled;
}
else
{
throw new NullReferenceException("A genetic operator was null.");
}
//note that each internal operator will adhere to the enabled flag itself,
//however the check is made here as this cannot be guaranteed with third party
//operators.
if (enabled)
{
genOp.Invoke(tempPopulation, ref processedPopulation, fitnessFunctionDelegate);
Evaluations += genOp.GetOperatorInvokedEvaluations();
tempPopulation.Solutions.Clear();
tempPopulation.Solutions.AddRange(processedPopulation.Solutions);
processedPopulation.Solutions.Clear();
// if the next operator in the collection does not require an evaluated population
// then no need to evaluate
if (op + 1 < operatorCount)
{
// we are not the last operator, there is another operator to follow
// so check to see if an evaluation is required
if (enabledOperators [op + 1].RequiresEvaluatedPopulation)
{
Evaluations += tempPopulation.Evaluate(fitnessFunctionDelegate);
}
else
{
EvaluationsRemoved++;
}
}
else
{
Evaluations += tempPopulation.Evaluate(fitnessFunctionDelegate);
}
//TODO: Add appropriate Logging
//if (this.OnLogging != null) {
// var eventArgs = new LoggingEventArgs ("Generation: {0}, Evaluations: {1}", currentGeneration + 1, Evaluations);
// this.OnLogging (this, eventArgs);
//}
//raise the Generation Complete event
if (this.OnOperatorComplete != null)
{
var eventArgs = new GaOperatorEventArgs(genOp, tempPopulation, currentGeneration + 1, Evaluations);
this.OnOperatorComplete(this, eventArgs);
}
}
}
return(tempPopulation);
}
internal Population RunGeneration(int currentGeneration, Population currentPopulation, FitnessFunction fitnessFunctionDelegate)
{
//create new empty populations for processing
var tempPopulation = currentPopulation.CreateEmptyCopy();
var processedPopulation = tempPopulation.CreateEmptyCopy();
tempPopulation.Solutions.AddRange(currentPopulation.Solutions);
//clear the current population, this keeps it simple as the solutions could
//easily get corrupted by genetic operators as from v 2.04 on, the genes are no longer cloned
currentPopulation.Solutions.Clear();
//invoke the operators
//Note: Each operator will contribute to populating the New Population. It is not untill all operators are
//invoked the the new population is complete. For example the Elite operator with 10% will return a
//_newPopulation object containing only the top 10% of the solutions that exist in _population.
//Deletions handled by the operator(s)
foreach (var op in this.Operators)
{
var enabled = true;
//maxFitnessPre = tempPopulation.MaximumFitness;
var genOp = op as IGeneticOperator; //this is the new interface
if (genOp != null)
{
//using new interface so check enabled
enabled = genOp.Enabled;
}
//note that each internal operator will adhere to the enabled flag itself,
//however the check is made here as this cannot be guaranteed with third party
//operators.
if (enabled)
{
op.Invoke(tempPopulation, ref processedPopulation, fitnessFunctionDelegate);
Evaluations += op.GetOperatorInvokedEvaluations();
tempPopulation.Solutions.Clear();
tempPopulation.Solutions.AddRange(processedPopulation.Solutions);
processedPopulation.Solutions.Clear();
//TODO: Fix this.
//Some solutions will have already been evaluated during the operator invocations
//e.g. crossover, however, they may have also been mutated. By tracking/evaluating
//mutated we could get here only needing to evaluate those that wern't touched
//by the operators. The Chromosome.EvaluatedByOperator flag could help here...
//evaluate in case it affects the next operators selection
Evaluations += tempPopulation.Evaluate(fitnessFunctionDelegate);
//raise the Generation Complete event
if (this.OnOperatorComplete != null)
{
var eventArgs = new GaOperatorEventArgs(op, tempPopulation, currentGeneration + 1, Evaluations);
this.OnOperatorComplete(this, eventArgs);
}
}
}
return(tempPopulation);
}
