/******TODO
*
* public Layer GetHLayer(int LayerIndex) => h_layers[LayerIndex];
* public void SetHLayer(int LayerIndex, Layer set) => h_layers[LayerIndex] = set;
*
* public void MutateAllLayer(float mutationProb)
* {
* for (int i = 0; i < h_layers.Length; i++)
* {
* h_layers[i].MutateAllNeurons(mutationProb);
* }
* }
*******/
public void Mutate(float mutationRate)
{
for (int i = 0; i < h_layers.CountLayer(); i++)
{
h_layers.Mutate(mutationRate);
}
}
public RNN(uint n_inputs, uint n_mem, uint[] shape, IActivation act)
{
n_memory = n_mem;
n_outputs = shape[shape.Length - 1];
mem_stack = new float[n_memory, n_outputs + n_inputs];
flat_mem_stack = Utils.FZerosArray(mem_stack);
activation = act;
//IActivation Tanh = new Tanh();
h_layers = new Sequential();
h_layers.Add(new Layer((uint)(n_inputs + flat_mem_stack.Length), shape[0], Activations.Tanh));
//h_layers[0] = new Layer((uint) (n_inputs + flat_mem_stack.Length), hidden_dim);
for (int i = 1; i < shape.Length; i++)
{
//h_layers[i] = new Layer(hidden_dim, hidden_dim);
h_layers.Add(new Layer(shape[i - 1], shape[i], Activations.Tanh));
}
// Setting the output layer activation to the given activation
Layer lastLayer = (Layer)h_layers.NeuralNetworks[h_layers.NeuralNetworks.Count - 1];
lastLayer.SetActivation(activation);
h_layers.NeuralNetworks[h_layers.CountLayer() - 1] = lastLayer;
this.n_outputs = shape[shape.Length - 1];
}
