Python 神经网络调教程序

01	import random

02	import math

03

04	from pyneurgen.neuralnet import NeuralNet

05	from pyneurgen.nodes import BiasNode, Connection

06

07	pop_len = 360

08	factor = 1.0 / float(pop_len)

09	population = [

10	(i, math.sin(float(i) * factor )) for i in range(pop_len)

11

]

12

13	all_inputs = []

14	all_targets = []

15

16	def population_gen(population):

17	pop_sort = [item for item in population]

18	random.shuffle(pop_sort)

19	for item in pop_sort:

20	yield item

21

22	#   Build the inputs

23	for position, target in population_gen(population):

24	pos = float(position)

25	all_inputs.append([random.random(), pos * factor])

26	all_targets.append([target])

27

28	net = NeuralNet()

29	net.init_layers(2, [10], 1)

30	net.randomize_network()

31	net.learnrate = .20

32

33	net.randomize_network()

34	net.set_all_inputs(all_inputs)

35	net.set_all_targets(all_targets)

36	length = len(all_inputs)

37

38	learn_end_point = int(length * .8)

39	net.set_learn_range(0, learn_end_point)

40	net.set_test_range(learn_end_point + 1, length - 1)

41	net.layers[1].set_activation_type('tanh')

42	net.learn(epochs=125, show_epoch_results=True,random_testing=False)

43	mse = net.test()

44

45	import matplotlib

46	from pylab import plot, legend, subplot, grid

47	from pylab import xlabel, ylabel, show, title

48

49	test_positions = [item[0][1] * 1000.0 for item in net.get_test_data()]

50

51	all_targets1 = [item[0][0] for item in net.test_actuals_targets]

52	allactuals = [item[1][0] for item in net.test_actuals_targets]

53

54	#   This is quick and dirty, but it will show the results

55	subplot(3, 1, 1)

56	plot([i[1] for i in population])

57	title("Population")

58	grid(True)

59

60	subplot(3, 1, 2)

61	plot(test_positions, all_targets1, 'bo', label='targets')

62	plot(test_positions, allactuals, 'ro', label='actuals')

63	grid(True)

64	legend(loc='lower left', numpoints=1)

65	title("Test Target Points vs Actual Points")

66

67	subplot(3, 1, 3)

68	plot(range(1, len(net.accum_mse) + 1, 1), net.accum_mse)

69	xlabel('epochs')

70	ylabel('mean squared error')

71	grid(True)

72	title("Mean Squared Error by Epoch")

73

74

show()