public void Learn()
{
List <int> usedInputs = new List <int>();
List <double> currentOutputs = new List <double>();
int era = 0;
do
{
usedInputs.Clear();
currentOutputs.Clear();
for (int i = 0; i < networkData.Inputs.Count; i++)
{
int currentInput;
do
{
currentInput = RandomHelper.GetRandomInRange(0, networkData.Inputs.Count);
} while (usedInputs.Contains(currentInput));
usedInputs.Add(currentInput);
int k = 0;
foreach (Neuron neuron in layers.getInputLayer().getNeurons())
{
neuron.setOutput(networkData.Inputs[currentInput][k++]);
}
sumWeightsAndActivate();
calculateSignalErrors(networkData.ExpectedOutputs[currentInput]);
updateWeights();
foreach (Neuron neuron in layers.getOutputLayer().getNeurons())
{
currentOutputs.Add(neuron.getOutput());
}
networkData.ActualOutputs.Add(currentInput, currentOutputs);
}
updateOverallError();
SingleEraEnded?.Invoke(new SingleEraData()
{
CurrentEra = era,
OverallError = networkData.OverallError,
LearnProgress = (era + 1d) / networkConfiguration.MaxEras,
PercentOfError = (networkConfiguration.MinError / networkData.OverallError * 100)
});
era++;
} while (era < networkConfiguration.MaxEras && networkData.OverallError > networkConfiguration.MinError);
}
public void Learn()
{
int era = 0;
Random r = new Random();
do
{
foreach (int currentInput in Enumerable.Range(0, networkData.Inputs.Count).OrderBy(x => r.Next()))
{
int k = 0;
foreach (Neuron neuron in layers.getInputLayer().getNeurons())
{
neuron.setOutput(networkData.Inputs[currentInput][k++]);
}
sumWeightsAndActivate();
calculateSignalErrors(networkData.ExpectedOutputs[currentInput]);
updateWeights();
List <double> currentOutputs = new List <double>();
foreach (Neuron neuron in layers.getOutputLayer().getNeurons())
{
currentOutputs.Add(neuron.getOutput());
}
networkData.ActualOutputs[currentInput] = new List <double>(currentOutputs);
}
updateOverallError();
SingleEraEnded?.Invoke(new SingleEraData()
{
CurrentEra = era,
OverallError = networkData.OverallError,
LearnProgress = (era + 1d) / networkConfiguration.MaxEras,
PercentOfError = (networkConfiguration.MinError / networkData.OverallError * 100)
});
era++;
} while (era < networkConfiguration.MaxEras && networkData.OverallError > networkConfiguration.MinError);
}