class BackPropagationLearningDialog(QDialog):
def __init__(self, network, test_data=None):
QDialog.__init__(self)
self.ui = Ui_LearningLauncher()
self.ui.setupUi(self)
self.network = network
self._setup_gui()
self.input_view = DataView(self.ui.learningDataView, header="Input", model=test_data,
columns=network.inputs_count() + len(network.get_layer(-1).get_neurons()))
# self.input_view.randomize_data()
@pyqtSlot()
def on_runButton_clicked(self):
error_metric = ErrorUtil.get_factor(str(self.ui.errorMetricSelector.currentText()))()
learning_factor = float(self.ui.learningFactorValue.text())
momentum = float(self.ui.momentumValue.text())
learning_iterations = int(self.ui.learningIterations.text())
bias_enabled = self.ui.biasEnabled.isChecked()
learning = BackpropagationLearning(self.network, error_metric, learning_factor, momentum, bias_enabled)
learning.learn(DataNormalizer.normalize(self.input_view.get_data()), learning_iterations)
self.accept()
@pyqtSlot()
def on_randomizeButton_clicked(self):
self.input_view.randomize_data()
@pyqtSlot()
def on_readInputButton_clicked(self):
filePath = QFileDialog().getOpenFileName(None, "Select input data file")
try:
with open(filePath, "r") as file:
file_lines = file.readlines()
rows_count = len(file_lines)
self.ui.dataCount.setValue(rows_count)
for (line, row) in zip(file_lines, range(rows_count)):
for (value, column) in zip(line.split(), range(self.input_view.column_count())):
self.input_view.set_data(row, column, float(value))
except Exception as e:
print e
@pyqtSlot(int)
def on_dataCount_valueChanged(self, value):
old_rows_count = self.input_view.row_count()
self.input_view.set_rows(value)
self.input_view.randomize_data(old_rows_count)
def _setup_gui(self):
self.ui.errorMetricSelector.addItems(ErrorUtil.registered_factors())
self.ui.errorMetricSelector.setCurrentIndex(self.ui.errorMetricSelector.findText(ErrorUtil.default_factor()))
def get_model(self):
return self.input_view.get_model()
class SimulationLauncherDialog(QDialog):
def __init__(self, network, test_data):
QDialog.__init__(self)
self.ui = Ui_SimulationLauncher()
self.ui.setupUi(self)
self.network = network
inputs_count = len(network.inputs)
self.input_view = DataView(self.ui.dataView, model=test_data, header="Input", columns=inputs_count)
outputs_count = len(network.layers[-1].neurons)
self.output_view = DataView(self.ui.outputView, header="Output", columns=outputs_count)
self.ui.testsCountSpinBox.setValue(self.input_view.row_count())
@pyqtSlot()
def on_runButton_clicked(self):
if not self.input_view.check_data():
QMessageBox.warning(None, "Error", "Input data is invalid")
else:
inputs = self.input_view.get_data()
for row_index in range(len(inputs)):
row = inputs[row_index]
response = self.network.calculate_network_response(row)
for i in range(len(response)):
self.output_view.set_data(row_index, i, round(response[i], 3))
@pyqtSlot()
def on_readFileButton_clicked(self):
filePath = QFileDialog().getOpenFileName(None, "Select input data file")
try:
with open(filePath, "r") as file:
file_lines = file.readlines()
rows_count = len(file_lines)
self.ui.testsCountSpinBox.setValue(rows_count)
for (line, row) in zip(file_lines, range(rows_count)):
for (value, column) in zip(line.split(), range(self.input_view.column_count())):
self.input_view.set_data(row, column, float(value))
except Exception as e:
print e
def get_model(self):
return self.input_view.get_model()
@pyqtSlot()
def on_generateRandomButton_clicked(self):
self.input_view.randomize_data()
@pyqtSlot(int)
def on_testsCountSpinBox_valueChanged(self, value):
self.input_view.set_rows(value)
self.output_view.set_rows(value)
class CounterPropagationLearningDialog(QDialog):
def __init__(self, network):
QDialog.__init__(self)
self.ui = Ui_LearningLauncher()
self.ui.setupUi(self)
self.network = network
self.network_layer = network.get_layer(0)
self.default_rectangular_topology_dimension = int(ceil(len(self.network_layer.get_neurons()) ** 0.5))
self._setup_gui()
self.input_view = DataView(
self.ui.learningDataView,
header="Input",
columns=network.inputs_count() + len(network.get_layer(-1).get_neurons()),
)
self.input_view.randomize_data()
@pyqtSlot()
def on_runButton_clicked(self):
if self.ui.neighbourhoodEnabled.isChecked():
neighbourhood = NeighbourhoodUtil.get_neighbourhood(
str(self.ui.neighbourhoodFunctionSelector.currentText())
)()
topology = TopologyUtil.get_topology(
str(self.ui.topologySelector.currentText()),
neighbourhood,
int(self.ui.initialNeighbourhoodRadius.text()),
int(self.ui.topologyRowsValue.text()),
int(self.ui.topologyColsValue.text()),
)
else:
topology = NoTopology()
self.metric = MetricUtil.get_metric(str(self.ui.metricSelector.currentText()))()
learning_factor = LearningFactorUtil.get_factor(
str(self.ui.kohonenLearningFactorSelector.currentText()),
float(self.ui.kohonenLearningFactorInitialValue.text()),
int(self.ui.kohonenLearningIterations.text()),
)
conscience_threshold = (
float(self.ui.conscienceThresholdValue.text()) if self.ui.conscienceEnabled.isChecked() else 0
)
kohonen_learning_iterations = int(self.ui.kohonenLearningIterations.text())
kohonen_learning = CounterPropagationKohonenLearning(
self.network_layer, topology, self.metric, learning_factor, conscience_threshold
)
learning_data = self.input_view.get_data()
grossberg_learning_rule = RulesUtil.get_rule(self.ui.learningRuleSelector.currentText())
grossberg_learning_factor = LearningFactorUtil.get_factor(
str(self.ui.grossbergLearningFactorSelector.currentText()),
float(self.ui.grossbergLearningFactorInitialValue.text()),
int(self.ui.grossbergLearningIterations.text()),
)
grossberg_learning_iterations = int(self.ui.grossbergLearningIterations.text())
cp_learning = CounterPropagationLearning(
self.network, grossberg_learning_rule, grossberg_learning_factor, kohonen_learning
)
cp_learning.learn(kohonen_learning_iterations, grossberg_learning_iterations, learning_data)
self.accept()
@pyqtSlot()
def on_randomizeButton_clicked(self):
self.input_view.randomize_data()
@pyqtSlot()
def on_readInputButton_clicked(self):
filePath = QFileDialog().getOpenFileName(None, "Select input data file")
try:
with open(filePath, "r") as file:
file_lines = file.readlines()
rows_count = len(file_lines)
self.ui.dataCount.setValue(rows_count)
for (line, row) in zip(file_lines, range(rows_count)):
for (value, column) in zip(line.split(), range(self.input_view.column_count())):
self.input_view.set_data(row, column, float(value))
except Exception as e:
print e
@pyqtSlot(int)
def on_dataCount_valueChanged(self, value):
old_rows_count = self.input_view.row_count()
self.input_view.set_rows(value)
self.input_view.randomize_data(old_rows_count)
@pyqtSlot(int)
def on_topologySelector_currentIndexChanged(self, value):
self.ui.topologyDimensionsBox.setVisible(self.ui.topologySelector.itemText(value).startsWith("2D"))
self.ui.topologyRowsValue.setText(str(self.default_rectangular_topology_dimension))
self.ui.topologyColsValue.setText(str(self.default_rectangular_topology_dimension))
@pyqtSlot()
def on_initializeWeightsButton_clicked(self):
kohonen_layer = self.network.get_layer(0)
weight = (1.0 / self.network.inputs_count()) ** 0.5
for neuron in range(len(kohonen_layer.get_neurons())):
for input in range(self.network.inputs_count()):
kohonen_layer.set_edge_weight(neuron, input, weight)
def _setup_gui(self):
self.ui.topologyDimensionsBox.hide()
self.ui.topologySelector.addItems(TopologyUtil.registered_topologies())
self.ui.topologySelector.setCurrentIndex(self.ui.topologySelector.findText(TopologyUtil.default_topology()))
self.ui.neighbourhoodFunctionSelector.addItems(NeighbourhoodUtil.registered_neighbourhoods())
self.ui.neighbourhoodFunctionSelector.setCurrentIndex(
self.ui.neighbourhoodFunctionSelector.findText(NeighbourhoodUtil.default_neighbourhood())
)
self.ui.metricSelector.addItems(MetricUtil.registered_metrics())
self.ui.metricSelector.setCurrentIndex(self.ui.metricSelector.findText(MetricUtil.default_metric()))
self.ui.kohonenLearningFactorSelector.addItems(LearningFactorUtil.registered_factors())
self.ui.kohonenLearningFactorSelector.setCurrentIndex(
self.ui.kohonenLearningFactorSelector.findText(LearningFactorUtil.default_factor())
)
self.ui.grossbergLearningFactorSelector.addItems(LearningFactorUtil.registered_factors())
self.ui.grossbergLearningFactorSelector.setCurrentIndex(
self.ui.grossbergLearningFactorSelector.findText(LearningFactorUtil.default_factor())
)
self.ui.learningRuleSelector.addItems(RulesUtil.rules())
self.ui.learningRuleSelector.setCurrentIndex(self.ui.learningRuleSelector.findText(RulesUtil.default_layer()))
self.ui.conscienceThresholdValue.setText(str(1 - round(float(1.0 / self.network.inputs_count()), 2) - 0.01))