internal static float[] GetDeltasHiddenLayer(float[] outputSignals, float[][] weightsNextLayer, float[] deltasErrorsNextLayer, ActivationFunc activationFunc)
{
int _neuronsCountThisLayer = outputSignals.Length;
int _neuronsCountNextLayer = deltasErrorsNextLayer.Length;
float[] _deltasErrors = new float[outputSignals.Length];
float _sumWeightsAndDeltasNextLayer;
float _derivativeActivFunc;
for (int _neuron = 0; _neuron < _neuronsCountThisLayer; _neuron++)
{
_sumWeightsAndDeltasNextLayer = GetSumWeightsAndDeltasNextLayer(_neuron);
_derivativeActivFunc = NeuralMath.GetDerivativeByFunction(outputSignals[_neuron], activationFunc);
_deltasErrors[_neuron] = _sumWeightsAndDeltasNextLayer * _derivativeActivFunc;
}
return(_deltasErrors);
float GetSumWeightsAndDeltasNextLayer(int neuronPrevLayer)
{
float _sumWeightsAndDeltas = 0;
for (int _neuronNextLayer = 0; _neuronNextLayer < _neuronsCountNextLayer; _neuronNextLayer++)
{
_sumWeightsAndDeltas += deltasErrorsNextLayer[_neuronNextLayer] * weightsNextLayer[neuronPrevLayer][_neuronNextLayer];
}
return(_sumWeightsAndDeltas);
}
}
private void UpdateMiniBatch(List <Tuple <double[, ], double[, ]> > miniBatch, double eta)
{
List <double[, ]> nablaB = new List <double[, ]>();
List <double[, ]> nablaW = new List <double[, ]>();
for (int i = 0; i < numberLayers - 1; i++)
{
nablaB.Add(new double[biases[i].GetLength(0), biases[i].GetLength(1)]);
nablaW.Add(new double[weights[i].GetLength(0), weights[i].GetLength(1)]);
}
foreach (Tuple <double[, ], double[, ]> batch in miniBatch)
{
Tuple <List <double[, ]>, List <double[, ]> > deltaNablas = BackPropogation(batch);
for (int j = 0; j < numberLayers - 1; j++)
{
nablaB[j] = NeuralMath.AddMatrix(nablaB[j], deltaNablas.Item1[j]);
nablaW[j] = NeuralMath.AddMatrix(nablaW[j], deltaNablas.Item2[j]);
}
}
for (int i = 0; i < numberLayers - 1; i++)
{
biases[i] = NeuralMath.SubtractMatrix(biases[i], NeuralMath.ScaleMatrix(nablaB[i], eta / miniBatch.Count));
weights[i] = NeuralMath.SubtractMatrix(weights[i], NeuralMath.ScaleMatrix(nablaW[i], eta / miniBatch.Count));
}
}
private double[,] FeedForward(double[,] a)
{
double[,] output = a;
for (int layer = 0; layer < numberLayers - 1; layer++)
{
double[,] wa = NeuralMath.DotMatrix(weights[layer], output);
output = NeuralMath.Sigmoid(NeuralMath.AddMatrix(wa, biases[layer]));
}
return(output);
}
internal static float[] GetDeltasOutputLayer(float[] outputSignals, float[] expectedSignals, ActivationFunc activationFunc)
{
int _neuronsCountThisLayer = outputSignals.Length;
float[] _deltasErrors = new float[_neuronsCountThisLayer];
float _errorOutSignal;
float _derivativeActivFunc;
for (int _neuron = 0; _neuron < _neuronsCountThisLayer; _neuron++)
{
_errorOutSignal = expectedSignals[_neuron] - outputSignals[_neuron];
_derivativeActivFunc = NeuralMath.GetDerivativeByFunction(outputSignals[_neuron], activationFunc);
_deltasErrors[_neuron] = _errorOutSignal * _derivativeActivFunc;
}
return(_deltasErrors);
}
private Tuple <List <double[, ]>, List <double[, ]> > BackPropogation(Tuple <double[, ], double[, ]> batch)
{
List <double[, ]> nablaB = new List <double[, ]>();
List <double[, ]> nablaW = new List <double[, ]>();
for (int i = 0; i < numberLayers - 1; i++)
{
nablaB.Add(new double[biases[i].GetLength(0), biases[i].GetLength(1)]);
nablaW.Add(new double[weights[i].GetLength(0), weights[i].GetLength(1)]);
}
// feed forward
double[,] activation = batch.Item1;
List <double[, ]> activations = new List <double[, ]> {
batch.Item1
};
List <double[, ]> zs = new List <double[, ]>();
for (int i = 0; i < numberLayers - 1; i++)
{
double[,] z = NeuralMath.AddMatrix(NeuralMath.DotMatrix(weights[i], activation), biases[i]);
zs.Add(z);
activation = NeuralMath.Sigmoid(z);
activations.Add(activation);
}
// backward pass
double[,] delta = NeuralMath.MultiplyMatrix(NeuralMath.CostDerivative(activations[activations.Count - 1], batch.Item2),
NeuralMath.SigmoidPrime(zs[zs.Count - 1]));
nablaB[nablaB.Count - 1] = delta;
nablaW[nablaW.Count - 1] = NeuralMath.DotMatrix(delta, NeuralMath.Transpose(activations[activations.Count - 2]));
for (int l = 2; l < numberLayers; l++)
{
double[,] z = zs[zs.Count - l];
double[,] sp = NeuralMath.SigmoidPrime(z);
delta = NeuralMath.MultiplyMatrix(NeuralMath.DotMatrix(
NeuralMath.Transpose(weights[weights.Count - l + 1]), delta), sp);
nablaB[nablaB.Count - l] = delta;
nablaW[nablaW.Count - l] = NeuralMath.DotMatrix(delta, NeuralMath.Transpose(activations[activations.Count - l - 1]));
}
return(Tuple.Create(nablaB, nablaW));
}
internal static float[] GetOutputSignalOfSample(float[] inputSignal, float[][] neuronWeights, ActivationFunc activationFunc, float[] isBiasNeuron)
{
int _synapsesCount = neuronWeights.Length;
int _neuronsCount = neuronWeights[0].Length;
float[] _outputSignals = new float[_neuronsCount];
float[][] _neuronWeightsT = TransposeArray(neuronWeights);
for (int neuron = 0; neuron < _neuronsCount; neuron++)
{
float _inputSignalForNeuron = 0;
for (int synapse = 0; synapse < _synapsesCount; synapse++)
{
_inputSignalForNeuron += inputSignal[synapse] * _neuronWeightsT[neuron][synapse] + (isBiasNeuron != null ? isBiasNeuron[neuron] : 0);
}
_outputSignals[neuron] = NeuralMath.GetApproximateByFunction(_inputSignalForNeuron, activationFunc);
}
return(_outputSignals);
}
internal static float[][] GetOutputSignalsOfAllSamples(float[][] inputSignals, float[][] neuronWeights, ActivationFunc activationFunc, float[] isBiasNeuron)
{
int _synapsesCount = neuronWeights.Length;
int _neuronsCount = neuronWeights[0].Length;
float[][] _outputSignals = new float[inputSignals.Length][];
float[][] _neuronWeightsT = TransposeArray(neuronWeights);
Parallel.For(0, inputSignals.Length, numberOfActiveDataset =>
{
float _inputSignalForNeuron;
if (_outputSignals[numberOfActiveDataset] == null)
{
_outputSignals[numberOfActiveDataset] = new float[_neuronsCount];
}
for (int neuron = 0; neuron < _neuronsCount; neuron++)
{
_inputSignalForNeuron = SumAllSynapseSignals(numberOfActiveDataset, neuron, _synapsesCount);
_outputSignals[numberOfActiveDataset][neuron] = NeuralMath.GetApproximateByFunction(_inputSignalForNeuron, activationFunc);
}
});
return(_outputSignals);
float SumAllSynapseSignals(int currentSet, int neuron, int synapsesOfNeuron)
{
float _inputSignal = 0;
for (int synapse = 0; synapse < synapsesOfNeuron; synapse++)
{
_inputSignal += inputSignals[currentSet][synapse] * _neuronWeightsT[neuron][synapse] + (isBiasNeuron != null ? isBiasNeuron[neuron] : 0);
}
return(_inputSignal);
}
}
