public void Initialize(CostFuntionKind costFunction, params int[] layersNodes)
{
if (initialized)
{
throw new System.InvalidOperationException();
}
if (layersNodes.Length < 2)
{
throw new System.ArgumentException();
}
inputSize = layersNodes[0];
outputSize = layersNodes[layersNodes.Length - 1];
neuralNetAccessor = new NeuralNetAccessor(layersNodes);
// Back propagation
this.costFunction = costFunction;
///
InitializeWeightsAndBiases();
//
initialized = true;
}
void InitializeLearning()
{
costFunctionGradient = new NeuralNetAccessor(neuralNetAccessor.InitializedParameters);
backPropagationOutput = new NeuralNetAccessor(neuralNetAccessor.InitializedParameters);
InitializeBackPropagation();
initializedLearning = true;
}
void AddRegularizationL1(NeuralNetAccessor costFunctionGradient, float regularizationRate)
{
for (int layer = 0; layer < costFunctionGradient.NumberOfLayers; layer++)
{
for (int node = 0; node < costFunctionGradient.NodesInLayer(layer); node++)
{
var costFunctionGradientWeights = costFunctionGradient.GetSigmoid(layer, node).Weights;
var neuralNetWeights = neuralNetAccessor.GetSigmoid(layer, node).Weights;
for (int i = 0; i < costFunctionGradientWeights.Length; i++)
{
costFunctionGradientWeights[i] += Mathf.Sign(neuralNetWeights[i]) * regularizationRate;
}
}
}
}
void AddRegularization(RegularizationMethodKind regularizationMethod, NeuralNetAccessor costFunctionGradient, float regularizationRate)
{
switch (regularizationMethod)
{
case RegularizationMethodKind.None:
break;
case RegularizationMethodKind.L2:
AddRegularizationL2(costFunctionGradient, regularizationRate);
break;
case RegularizationMethodKind.L1:
AddRegularizationL1(costFunctionGradient, regularizationRate);
break;
default:
throw new System.NotImplementedException();
}
}
public BackPropagationQuadratic(NeuralNetAccessor neuralNetAccessor)
{
this.neuralNetAccessor = neuralNetAccessor;
Initialize();
}
public BackPropagationCrossEntropy(NeuralNetAccessor neuralNetAccessor)
{
this.neuralNetAccessor = neuralNetAccessor;
Initialize();
}