private void Feed_Forward(double[] input)
{
Activation_Function af = new Activation_Function();
for (int i = 0; i < layer.Count; i++)
{
if (i == 0)
{
for (int j = 0; j < layer[i].neuron.Count; j++)
{
layer[i].neuron[j].output = input[j];
}
}
else
{
for (int j = 0; j < layer[i].neuron.Count; j++)
{
double weighted_sum = 0;
for (int k = 0; k < layer[i - 1].neuron.Count; k++)
{
weighted_sum += layer[i - 1].neuron[k].output * layer[i - 1].neuron[k].weight[j];
}
layer[i].neuron[j].output = af.hyperbolic_tangent(weighted_sum + layer[i - 1].bias.output);
}
}
}
}
private void Back_Propagate(double[] input)
{
Activation_Function af = new Activation_Function();
for (int i = layer.Count; i > 0; i--)
{
if (i == layer.Count)
{
for (int j = 0; j < layer[i - 1].neuron.Count; j++)
{
layer[i - 1].neuron[j].error = (input[j] - layer[i - 1].neuron[j].output) * af.hyperbolic_tangent_derivative(layer[i - 1].neuron[j].output);
}
}
else
{
for (int j = 0; j < layer[i - 1].neuron.Count; j++)
{
double error_weight = 0;
for (int k = 0; k < layer[i].neuron.Count; k++)
{
error_weight += layer[i - 1].neuron[j].weight[k] * layer[i].neuron[k].error;
}
layer[i - 1].neuron[j].error = error_weight * af.hyperbolic_tangent_derivative(layer[i - 1].neuron[j].output);
}
}
}
}
