public static extern int imaqLearnPattern3(IntPtr image, LearningMode learningMode, ref LearnPatternAdvancedOptions advancedOptions, IntPtr mask);
public virtual void EvolveTimingChromosome(LearningMode learningMode, Dictionary<BitArray,double> agents,
double crossOverProbability, double mutationProbability)
{
switch (learningMode)
{
case LearningMode.Random:
RandomizeTimingChromosome(_random);
break;
case LearningMode.GA:
if (_random.NextDouble() < crossOverProbability)
{
//selection
var chromosomes =
agents.Select(a => new {a.Key, Rank = a.Value*_random.NextDouble()});
var chosenChromosome = chromosomes.OrderByDescending(c => c.Rank).First();
//crossover
var crossOver = Math.Floor(_random.NextDouble()*TimingChromosome.Count);
for (int i = 0; i < crossOver; i++)
{
TimingChromosome[i] = chosenChromosome.Key[i];
}
}
//mutation
for (int i = 0; i < TimingChromosome.Count; i++)
{
TimingChromosome[i] = _random.NextDouble() < mutationProbability
? !TimingChromosome[i]
: TimingChromosome[i];
}
break;
case LearningMode.MRE:
//update propensities
for (int i = 0; i < _learningLog.Count; i++)
{
double reward;
if (_learningLog[i].TimingChromosome == TimingChromosome)
{
reward = CurrentProfit*(1 - _experimentation);
}
else
{
reward = _learningLog[i].Propensity * (_experimentation / (_learningLog.Count -1));
}
_learningLog[i].Propensity = (1 - _recency)*_learningLog[i].Propensity + reward;
}
//update probabilities
for (int i = 0; i < _learningLog.Count; i++)
{
_learningLog[i].Probability = Math.Exp(_learningLog[i].Propensity/_temperature)/
_learningLog.Sum(l => Math.Exp(l.Propensity/_temperature));
}
//select chromosome
TimingChromosome = _learningLog.OrderByDescending(l => l.Probability * _random.NextDouble()).First().TimingChromosome;
break;
default:
throw new ArgumentOutOfRangeException("Unknown learning mode");
}
CurrentProfit = 0;
}
public static extern int imaqLearnPattern(IntPtr image, LearningMode learningMode);
