private TrainingParameters GetParameters()
{
int epochs = Int32.Parse(TrainingEpochs.Text);
int iterations = Int32.Parse(TrainingIterations.Text);
TrainingParameters p = p = new KMeansTrainingParameters()
{
Epochs = epochs,
TrainingData = TrainingData
};
if (Network is SelfOrganizingMap)
{
int kmax = TrainingData.Count * epochs;
double minLearningRate = double.Parse(TrainingLearningRateMin.Text, CultureInfo.InvariantCulture);
double maxLearningRate = double.Parse(TrainingLearningRateMax.Text, CultureInfo.InvariantCulture);
p = new SomTrainingParameters()
{
LearningRate = new DeclineExponentially(kmax, minLearningRate, maxLearningRate),
TirednessMechanism = GetTiredMechanism(),
NeighbourhoodFunction = GetNeighbourhoodFunction(kmax),
NumberOfIterations = iterations,
TrainingData = TrainingData
};
}
return(p);
}
public override void Train(TrainingParameters parameters)
{
KMeansTrainingParameters p = (KMeansTrainingParameters)parameters;
try
{
p.Validate();
}
catch (TrainingParametersException e)
{
throw new SomLogicalException($"Improper training parameters. {e.Message}");
}
var trainingData = parameters.TrainingData;
var epochs = p.Epochs;
if (trainingData.First().Count != Dimensions)
{
throw new SomLogicalException("Training data points dimensions doesn't match som dimensions.");
}
Observer?.SaveState(Neurons);
bool end = false;
for (int j = 0; j < epochs && !end; j++)
{
var shuffled = trainingData.Shuffle();
Dictionary <Neuron, List <Vector <double> > > neuronsZones = new Dictionary <Neuron, List <Vector <double> > >();
foreach (var neuron in Neurons)
{
neuronsZones.Add(neuron, new List <Vector <double> >());
}
for (int i = 0; i < shuffled.Count; i++)
{
Dictionary <Neuron, double> neuronsDistances = new Dictionary <Neuron, double>();
for (int k = 0; k < Neurons.Count; k++)
{
//store neurons with their distances
neuronsDistances.Add(Neurons[k],
DistanceCalculator.CalculateDistance(shuffled[i], Neurons[k].CurrentWeights));
}
Neuron winner = neuronsDistances.OrderBy(pair => pair.Value).First().Key;
neuronsZones[winner].Add(shuffled[i]);
}
var aggregate = Vector <double> .Build.DenseOfEnumerable(new List <double>(new double[Dimensions]));
end = true;
foreach (var pair in neuronsZones)
{
foreach (var vec in pair.Value)
{
aggregate = aggregate.Add(vec);
}
if (pair.Value.Count != 0)
{
aggregate = aggregate.Divide(pair.Value.Count);
var diff = aggregate.Subtract(pair.Key.CurrentWeights).Map(Math.Abs).Sum();
if (end && diff > 0.0001)
{
end = false;
}
pair.Key.CurrentWeights = aggregate;
}
}
Observer?.SaveState(Neurons);
}
}
