public Vector ProcessInput(IFunctionApproximator network,
Vector input)
{
hiddenLayer.FeedForward(input);
outputLayer.FeedForward(hiddenLayer.GetLastActivationValues());
return(outputLayer.GetLastActivationValues());
}
public void SetNeuralNetwork(IFunctionApproximator fapp)
{
FeedForwardDeepNeuralNetwork ffnn = fapp as FeedForwardDeepNeuralNetwork;
if (null == ffnn)
{
throw new Exception("Only supporting FeedForwardDeepNeuralNetwork at this stage.");
}
this.hiddenLayers = new Layer[ffnn.GetNumberOfHiddenLayers()];
for (int i = 0; i < hiddenLayers.Length; ++i)
{
this.hiddenLayers[i] = ffnn.GetHiddenLayer(i);
}
this.outputLayer = ffnn.GetOutputLayer();
this.hiddenSensitivities = new LayerSensitivity[ffnn.GetNumberOfHiddenLayers()];
for (int i = 0; i < hiddenLayers.Length; ++i)
{
this.hiddenSensitivities[i] = new LayerSensitivity(ffnn.GetHiddenLayer(i));
}
this.outputSensitivity = new LayerSensitivity(outputLayer);
}
public void SetNeuralNetwork(IFunctionApproximator fapp)
{
FeedForwardNeuralNetwork ffnn = (FeedForwardNeuralNetwork)fapp;
this.hiddenLayer = ffnn.GetHiddenLayer();
this.outputLayer = ffnn.GetOutputLayer();
this.hiddenSensitivity = new LayerSensitivity(hiddenLayer);
this.outputSensitivity = new LayerSensitivity(outputLayer);
}
public Vector ProcessInput(IFunctionApproximator network, Vector input)
{
hiddenLayers[0].FeedForward(input);
for (int i = 1; i < hiddenLayers.Length; ++i)
{
hiddenLayers[i].FeedForward(hiddenLayers[i - 1].GetLastActivationValues());
}
outputLayer.FeedForward(hiddenLayers[hiddenLayers.Length - 1].GetLastActivationValues());
return(outputLayer.GetLastActivationValues());
}
public void ProcessError(IFunctionApproximator network,
Vector error)
{
// TODO calculate total error somewhere
// create Sensitivity Matrices
outputSensitivity.SensitivityMatrixFromErrorMatrix(error);
hiddenSensitivities[hiddenSensitivities.Length - 1].SensitivityMatrixFromSucceedingLayer(outputSensitivity);
for (int i = hiddenSensitivities.Length - 2; i >= 0; --i)
{
hiddenSensitivities[i].SensitivityMatrixFromSucceedingLayer(hiddenSensitivities[i + 1]);
}
// calculate weight Updates
CalculateWeightUpdates(outputSensitivity, hiddenLayers[hiddenLayers.Length - 1].GetLastActivationValues(),
learningRate, momentum);
for (int i = hiddenLayers.Length - 1; i > 0; --i)
{
CalculateWeightUpdates(hiddenSensitivities[i], hiddenLayers[i - 1].GetLastActivationValues(),
learningRate, momentum);
}
CalculateWeightUpdates(hiddenSensitivities[0], hiddenLayers[0].GetLastInputValues(), learningRate, momentum);
// calculate Bias Updates
CalculateBiasUpdates(outputSensitivity, learningRate, momentum);
for (int i = hiddenLayers.Length - 1; i >= 0; --i)
{
CalculateBiasUpdates(hiddenSensitivities[i], learningRate, momentum);
}
// update weightsAndBiases
outputLayer.UpdateWeights();
outputLayer.UpdateBiases();
for (int i = hiddenLayers.Length - 1; i >= 0; --i)
{
hiddenLayers[i].UpdateWeights();
hiddenLayers[i].UpdateBiases();
}
}
public void ProcessError(IFunctionApproximator network,
Vector error)
{
// TODO calculate total error somewhere
// create Sensitivity Matrices
outputSensitivity.SensitivityMatrixFromErrorMatrix(error);
hiddenSensitivity.SensitivityMatrixFromSucceedingLayer(outputSensitivity);
// calculate weight Updates
CalculateWeightUpdates(outputSensitivity, hiddenLayer.GetLastActivationValues(), learningRate, momentum);
CalculateWeightUpdates(hiddenSensitivity, hiddenLayer.GetLastInputValues(), learningRate, momentum);
// calculate Bias Updates
CalculateBiasUpdates(outputSensitivity, learningRate, momentum);
CalculateBiasUpdates(hiddenSensitivity, learningRate, momentum);
// update weightsAndBiases
outputLayer.UpdateWeights();
outputLayer.UpdateBiases();
hiddenLayer.UpdateWeights();
hiddenLayer.UpdateBiases();
}
