def updateSeriesTable(self):
self.Output_Series.clear()
selectedDataStats = self.data.getSelectedDataStats()
self.Output_Series.setRowCount(len(selectedDataStats))
self.Output_Series.setColumnCount(2)
for row, val in enumerate(selectedDataStats):
print(lstr(row + 0.5))
label, value = val
title = QtWidgets.QTableWidgetItem(label)
self.Output_Series.setItem(row, 0, title)
if not isinstance(value, str):
value = lstr(value)
print(value)
value = QtWidgets.QTableWidgetItem(value)
self.Output_Series.setItem(row, 1, value)
self.Output_Series.resizeColumnsToContents()
if i >= cycles:
break
return errors_data
funs_data = []
for f in funcs:
print(f.__name__)
funs_data.append(train(f))
return funs_data
times = 10
cycles = 100
# repeat for average
avg_data = train_network()
for _ in range(times - 1):
data = train_network()
# consolidate
for i, new_func_data, avg_func_data in zip(range(len(funcs)), data,
avg_data):
avg_data[i] = [sum(d) for d in zip(new_func_data, avg_func_data)]
# divide and convert to polish notation (use locale)
for i, func_data in enumerate(avg_data):
avg_data[i] = [lstr(error / times) for error in func_data]
# transpose for excel
avg_data = list(zip(*avg_data))
# save
with open("neural_data.csv", 'w') as file:
writer = csv.writer(file, delimiter=';') # delimit with semicolons
writer.writerows(avg_data)
print(f"Generation: #{generation}")
# evaluate for fitness
evaluate_networks(networks)
# sort with fittest at the top
networks.sort(key=lambda n: n.fitness, reverse=True)
best_net = networks[0]
# save the best net if needed
print(f"Best fitness: {best_net.fitness}")
if best_net.fitness > best_net_data[0]:
best_net_data = (best_net.fitness, best_net.export_data())
print(f"New best: {best_net.fitness}\n")
snake_data.append([
generation, # current generation
best_net_data[0], # current best fitness
# average fitness of a population
lstr(sum(n.fitness for n in networks) / len(networks)),
])
if generation >= max_generation:
break
# select for the next generation
networks = select_networks(networks)
# crossover for the next generation
networks = crossover_networks(networks)
# save data for the performance graph
with open("snake_data.csv", 'w') as file:
writer = csv.writer(file, delimiter=';') # delimit with semicolons
writer.writerows(snake_data)
# restore the best network
best_net = Network(net_args)
best_net.fitness = best_net_data[0]
def setInitialUIvalues(self):
self.NeutralMass.setText(lstr(22870))
self.MassAccuracy.setText(lstr(200))
self.MinCS.setText(lstr(1))
self.MaxCS.setText(lstr(50))
self.Calibration_a.setText(lstr(231.7))
self.Calibration_b.setText(lstr(118.7))
self.Calibration_X.setText(lstr(0.6262))
self.TransferParam.setText(lstr(1.41))
self.PusherDelay.setText(lstr(110))
self.Gas_mass.setText(lstr(28))
print(lstr(255.3))