def dialog(network, learning_data):
layers = network.get_layers()
if Layers.get_type(layers[0]) == Layers.KOHONEN_LAYER:
if len(layers) == 2 and Layers.get_type(layers[1]) == Layers.GENERAL_LAYER:
return CounterPropagationLearningDialog(network)
else:
return KohonenLearningDialog(network)
else:
return BackPropagationLearningDialog(network, learning_data)
def _setup_network_parameters(self):
self.ui.layerTypeSelector.addItems(Layers.layers_types())
self.ui.layerTypeSelector.setCurrentIndex(
self.ui.layerTypeSelector.findText(Layers.default_layer())
)
self.ui.layersEdit.setValue(len(self.network_model.layers()))
for layer in self.network_model.layers():
self.ui.layerComboBox.addItem(layer.layer_name())
self.ui.inputNeuronsEdit.setValue(self.network_model.network_inputs())
self.ui.neuronsEdit.setValue(self.network_model.default_layer().rowCount())
def _create_layer(self, standard_layer_data):
layer = Layers.create_layer(standard_layer_data.layer_type())
layer_data = standard_layer_data.get_neurons_data()
for neuron_data in layer_data:
layer.add_neuron(Neuron(
PSPUtil.get_psp(neuron_data[0]),
ActivatorUtil.get_activator(neuron_data[1]),
float(neuron_data[2]))
)
return layer
def add_layer(self):
layer_name = self._layer_name(len(self.layers_data) + 1)
input_neurons = self.input_neurons
if len(self.layers_data) != 0:
input_neurons = self.layers_data[-1].rowCount()
layer_model = LayerModel(layer_name, Layers.default_layer(), NetworkModel.DEFAULT_NEURONS_COUNT, input_neurons,
self.minWeight,
self.maxWeight)
self.layers_data.append(layer_model)
return layer_model
def from_network(network):
model = NetworkModel(empty=True)
model.set_network_input_neurons(network.inputs_count())
for layer in network.get_layers():
layer_model = model.add_layer()
neurons = layer.get_neurons()
layer_model.set_layer_type(Layers.get_type(layer))
layer_model.set_neurons(len(neurons))
for neuron_index in range(len(neurons)):
neuron = neurons[neuron_index]
layer_model.set_psp_for_neuron(neuron_index, neuron.get_psp().NAME)
layer_model.set_activation_for_neuron(neuron_index, neuron.get_activator().NAME)
layer_model.set_bias_for_neuron(neuron_index, neuron.get_bias().get_weight())
inputs = neuron.get_inputs()
for input_index in range(len(inputs)):
layer_model.set_weight_for_connection(neuron_index, input_index,
round(neuron.get_input_edge(input_index).get_weight(), 3))
return model
