def calculate(self):
"""Calculates semantics of all hidden neurons and output neuron."""
i = 0
for layer in self.cnn_layers:
if i == 0:
for neuron in layer:
neuron.calculate()
else:
for neuron in layer:
neuron.calculate2()
i += 1
last_layer = self.cnn_layers[-1]
flatten_layer = None
for neuron in last_layer:
temp_arr = neuron.semantics
temp_arr2 = temp_arr.reshape(-1, neuron.semantics.shape[3])
if flatten_layer is not None:
flatten_layer = concatenate((flatten_layer, temp_arr2), axis=0)
else:
flatten_layer = temp_arr2
self.flatten_layer = array([Sensor(d) for d in flatten_layer])
#todo tak zupelnie nie moze byc liczba innput_connections powinna byc dokladnie idealna !!!! cos tu nie gra do naprawy!!!
self._connect_flat_hidden(self.flatten_layer, self.hidden_layers[0])
self.load_semantic_after_flat_layer()
for hidden_layer in self.hidden_layers:
for neuron in hidden_layer:
neuron.calculate()
for neuron in self.output_layer:
neuron.calculate()
def create_network_from_topology(topology):
"""Creates neural network from topology."""
# Create bias.
bias = Sensor(array([1]))
# Creates neurons from remaining items in string.
activation_function = choice(list(_NON_LINEAR_ACTIVATION_FUNCTIONS.keys()))
hidden_layers = [[create_neuron(activation_function, bias) for i in range(j)] for j in topology]
# Create output neuron.
output_neuron = create_neuron('identity', bias)
# Connect nodes in neural network.
if len(hidden_layers) > 1:
for i in range(1, len(hidden_layers)): _connect_nodes(hidden_layers[i - 1], hidden_layers[i])
_connect_nodes(hidden_layers[i], [output_neuron])
else:
_connect_nodes(hidden_layers[0], [output_neuron])
# Create neural network.
neural_network = NeuralNetwork(None, bias, hidden_layers, output_neuron)
# Return neural network.
return neural_network
def create_network_from_topology(topology):
"""Creates neural network from topology."""
# Create bias.
bias = Sensor(array([1]))
# Creates neurons from remaining items in string.
hidden_layers = [[create_neuron('tanh', bias) for i in range(j)]
for j in topology]
# Create output neuron.
output_neuron = create_neuron('identity', bias)
# Connect nodes in neural network.
if len(hidden_layers) > 1:
for i in range(1, len(hidden_layers)):
_connect_nodes(hidden_layers[i - 1], hidden_layers[i])
_connect_nodes(hidden_layers[i], [output_neuron])
else:
_connect_nodes(hidden_layers[0], [output_neuron])
# Create neural network.
neural_network = NeuralNetwork(None, bias, hidden_layers, output_neuron)
# Return neural network.
return neural_network
def _initialize_bias(self, neural_network):
"""Initializes biases with same length as sensors."""
return Sensor(
resize(array([1]), shape(neural_network.sensors[0].semantics)))
def _initialize_flat_sensors(self):
"""Initializes sensors based on input matrix."""
return [Sensor(input_data) for input_data in self.input_matrix.T]
def add_sensors(self, X):
"""Adds sensors to neural network and loads input data."""
self.sensors = [Sensor(np.array([])) for sensor_data in X.T]
self.load_sensors(X)
# Connects nodes to first level hidden layer.
_connect_nodes(self.sensors, self.hidden_layers[0])
def _initialize_flat_sensors(self, data):
"""Initializes sensors based on input matrix."""
return [Sensor(d) for d in data]
def _initialize_bias(self, neural_network):
"""Initializes biases with same length as sensors."""
return Sensor(ones(neural_network.sensors[0].semantics.shape))