public NeuralNetwork(params int[] units)
{
learning_rate = 0.03f;
momentum_rate = 0.3f;
error_limit = 0.005f;
ignore_error = false;
infinite_loop = false;
max_circle = int.MaxValue;
circle = 0;
score = 0f;
fitness = 0f;
id = 0;
layers = new Layer[units.Length];
for (int i = 0; i < layers.Length; i++)
{
PerceptronType.Type type = PerceptronType.Type.hidden;
if (i == 0)
{
type = PerceptronType.Type.input;
}
else if (i == layers.Length - 1)
{
type = PerceptronType.Type.output;
}
layers[i] = new Layer(units[i], (i != 0), ActivationType.Type.sigmoid, type);
}
input_layer = this.layers[0];
hidden_layers = new Layer[layers.Length - 2];
for (int i = 0; i < hidden_layers.Length; i++)
{
hidden_layers[i] = this.layers[i + 1];
}
output_layer = this.layers[this.layers.Length - 1];
for (int i = 0; i < layers.Length - 1; i++)
{
Layer.Connect(this.layers[i], this.layers[i + 1]);
}
}
public Layer(int len, bool bias_enable, ActivationType.Type activation_type, PerceptronType.Type perceptron_type)
{
type = perceptron_type;
for (int i = 0; i < len; i++)
{
perceptrons.Add(new Perceptron()
{
activation_type = activation_type,
type = type,
tag = "untitled",
current_layer = this
});
}
if (bias_enable)
{
bias = new Perceptron()
{
type = PerceptronType.Type.bias,
tag = "bias",
state = 1f,
current_layer = this
};
}
}
