protected virtual void OnBeginSample(int currentIteration, CY_NN_Dataset_Entry e)
{
if (BeginSampleEvent != null)
{
BeginSampleEvent(this, new CY_Dataset_Entry_Args(currentIteration, e));
}
}
public virtual void Learn(CY_NN_Dataset set, int trainingEpochs)
{
isStopping = false;
Initialize();
for (int currentIteration = 0; currentIteration < trainingEpochs; currentIteration++)
{
int[] randomOrder = GetRandomOrder(set.trainingSamples.Count);
OnBeginEpoch(currentIteration, set);
if (jitterEpoch > 0 && currentIteration % jitterEpoch == 0)
{
for (int i = 0; i < connectors.Count; i++)
{
connectors[i].noise(jitterNoiseLimit);
}
}
for (int index = 0; index < set.trainingSamples.Count; index++)
{
CY_NN_Dataset_Entry e = set[randomOrder[index]];
OnBeginSample(currentIteration, e);
LearnSample(set[randomOrder[index]], currentIteration, trainingEpochs);
OnEndSample(currentIteration, e);
if (isStopping)
{
isStopping = false; return;
}
}
OnEndEpoch(currentIteration, set);
if (isStopping)
{
isStopping = false; return;
}
}
}
protected override void LearnSample(CY_NN_Dataset_Entry e, int currentIteration, int trainingEpochs)
{
int layerCount = layers.Count;
inputLayer.SetInput(e.inputV);
for (int i = 0; i < layerCount; i++)
{
layers[i].calcFire();
}
mse += (outputLayer as CY_NeuronLayer).SetErrors(e.outputV);
for (int i = layerCount; i > 0;)
{
CY_NeuronLayer layer = layers[--i] as CY_NeuronLayer;
if (layer != null)
{
layer.calcError();
}
}
for (int i = 0; i < layerCount; i++)
{
layers[i].Learn(currentIteration, trainingEpochs);
}
}
public CY_Dataset_Entry_Args(int i, CY_NN_Dataset_Entry e)
{
this.i = i;
this.e = e;
}
public override void Learn(CY_NN_Dataset_Entry e, int i, int max_i)
{
mse = 0d;
base.Learn(e, i, max_i);
}
protected abstract void LearnSample(CY_NN_Dataset_Entry e, int i, int max_i);
public virtual void Learn(CY_NN_Dataset_Entry e, int i, int max_i)
{
OnBeginSample(i, e);
LearnSample(e, i, max_i);
OnEndSample(i, e);
}