def testForProperLearningRateForNeuron(self, mocked_print):
learning_rate = 0.01
net = NeuralNet(2,
3,
2,
1, [1, 2], [1], [1, 2, 3, 4, 5, 6],
learning_rate=learning_rate)
self.assertEqual(net.get_output_layer()[0].get_learning_rate(),
learning_rate)
def test_four_neurons_for_not(self, mocked_print):
x_test = np.array([[1], [0]])
y_test = np.array([[0], [1]])
net = NeuralNet(1, 3, 2)
net.train(x_test, y_test)
sys.stderr.write(f'\n\nFour neuron net predicting for NOT:\n')
for test, expected in zip([[1], [0], [1], [0]], [0, 1, 0, 1]):
test_result = net.predict(test)
sys.stderr.write(
f'Input: {test}, Expected: {expected}, Prediction: {test_result}\n'
)
self.assertTrue(np.abs(test_result - expected) < 0.1)
def test_three_neurons_for_equals(self, mocked_print):
x_test = np.array([[1], [0]])
y_test = np.array([[1], [0]])
net = NeuralNet(1, 3, 1)
net.train(x_test, y_test)
sys.stderr.write(f'\n\nThree neuron net predicting for EQUALS:\n')
for test, expected in zip([[1], [0], [1], [0]], [1, 0, 1, 0]):
test_result = net.predict(test)
sys.stderr.write(
f'Input: {test}, Expected: {round(expected)}, Prediction: {test_result}\n'
)
self.assertTrue(np.abs(test_result - expected) < 0.1)
def test_seven_neurons_for_mod2(self, mocked_print):
x_test = np.array([[1, 0], [0, 0], [0, 1], [1, 1]])
y_test = np.array([[1], [0], [1], [0]])
net = NeuralNet(2, 3, 4)
net.train(x_test, y_test, 20000)
sys.stderr.write(f'\n\nSeven neuron net predicting for MOD 2:\n')
for test, expected in zip([[1, 0], [0, 1], [1, 1], [0, 0]],
[1, 1, 0, 0]):
test_result = net.predict(test)
sys.stderr.write(
f'Input: {test}, Expected: {expected}, Prediction: {test_result}\n'
)
self.assertTrue(np.abs(test_result - expected) < 0.1)
def testForProperActivationFunction(self, mocked_print):
def fake_function():
pass
with self.assertRaises(TypeError):
NeuralNet(2, 3, 2, 1, [1, 2], [1], [1, 2, 3, 4, 5, 6],
fake_function)
def setUp(self):
self.patcher = patch('builtins.print')
self.mock_object = self.patcher.start()
self.x_test = np.array([[1, 0], [0, 1], [0, 0], [1, 1]])
self.y_test = np.array([[0, 1], [0, 1], [1, 0], [1, 0]])
self.hidden_node_biases = [2, 2]
self.output_node_biases = [3, 3]
self.weights = [5, 5, 5, 5, 5, 5, 5, 5]
self.net = NeuralNet(2,
3,
2,
2,
self.hidden_node_biases,
self.output_node_biases,
self.weights,
activation_function=ActivationFunction.LOG,
learning_rate=0.01)
def setUp(self):
self.patcher = patch('builtins.print')
self.mock_object = self.patcher.start()
self.x_test = np.array([[1], [0]])
self.y_test = np.array([[1], [0]])
self.hidden_node_biases = [0.10399593413625441, 0.9477095968224993]
self.output_node_biases = [0.1683069001902736]
self.weights = [
0.08946817286422626, 0.10066428298027419, 0.8768683269443115
]
self.net = NeuralNet(1,
4,
1,
1,
self.hidden_node_biases,
self.output_node_biases,
self.weights,
learning_rate=0.01)
def test_run_net(self, mocked_print):
number_of_input_neurons = 54
nodes_per_layer = 50
number_of_layers = 3
number_of_output_neurons = 27
number_of_epochs = 1
learning_rate = .3
word_file = open('../words.txt', 'r')
words = []
for w in word_file.readlines():
words.append(w)
word_file.close()
sys.stderr.write(f'Input of {len(words)} words.')
converter = WordConverter(words)
x_test, y_test = converter.get_input()
net = NeuralNet(number_of_input_neurons=number_of_input_neurons,
number_of_layers=number_of_layers,
number_of_neurons_per_layer=nodes_per_layer,
number_of_output_neurons=number_of_output_neurons,
activation_function=ActivationFunction.LOG,
learning_rate=learning_rate)
start_time = time.time()
sys.stderr.write(f'\nTraining {nodes_per_layer} nodes per layer with '
f'{number_of_layers - 2 } hidden layers for '
f'{number_of_epochs} times.')
net.train(np.asarray(x_test), np.asarray(y_test), number_of_epochs)
sys.stderr.write(
f'\nTraining took {time.time() - start_time} seconds.\n')
test_input = x_test[:50]
test_output = y_test[:50]
for test, expected in zip(test_input, test_output):
test_result = net.predict(test)
letter_1 = converter.letter_dict(list(test[:27]).index(1))
letter_2 = converter.letter_dict(list(test[27:]).index(1))
sys.stderr.write(
f'Input: {letter_1+letter_2}, '
f'Expected: {converter.letter_dict(list(expected).index(1))}, '
f'Prediction: {converter.letter_dict(test_result)}\n')
self.assertTrue(True)
def test_three_inputs_three_outputs_and_three_hidden_for_mod2(
self, mocked_print):
x_test = np.array([[1, 0, 1], [0, 1, 0], [0, 1, 1], [1, 1, 0],
[1, 0, 0]])
y_test = np.array([[0, 0, 1], [0, 1, 0], [1, 0, 0], [0, 1, 0],
[1, 0, 0]])
net = NeuralNet(3, 3, 3, 3, activation_function=ActivationFunction.LOG)
sys.stderr.write(f'\n\n------------------TRAINING------------------')
net.train(x_test, y_test, 100000)
sys.stderr.write(
f'\n\nNine neuron net (3 input, 3 hidden, 3 output) predicting for MOD 2:\n'
)
test_input = [[1, 0, 0], [0, 1, 0], [1, 1, 0], [0, 0, 1]]
# test_output = [[1, 0, 0], [0, 1, 0], [0, 1, 0], [0, 0, 1]]
test_output = [0, 1, 1, 2]
for test, expected in zip(test_input, test_output):
test_result = net.predict(test)
sys.stderr.write(
f'Input: {test}, Expected: {expected}, Prediction: {test_result}\n'
)
self.assertTrue(np.abs(test_result - expected) < 0.1)
def test_run_net(self, mocked_print):
for number_of_nodes_per_layer in [6, 9, 15]:
for number_of_layers in range(3, 5):
for number_of_epochs in [50000, 100000]:
x_test = np.array([[1, 0, 0, 1, 0, 0], [1, 0, 0, 0, 1, 0],
[1, 0, 0, 0, 0, 1], [0, 1, 0, 1, 0, 0],
[0, 1, 0, 0, 1, 0], [0, 1, 0, 0, 0, 1],
[0, 0, 1, 1, 0, 0], [0, 0, 1, 0, 1, 0],
[0, 0, 1, 0, 0, 1]])
y_test = np.array([[1, 0, 0], [0, 1, 0], [0, 0, 1],
[0, 1, 0], [0, 0, 1], [1, 0, 0],
[0, 0, 1], [1, 0, 0], [0, 1, 0]])
net = NeuralNet(6,
number_of_layers,
number_of_nodes_per_layer,
3,
activation_function=ActivationFunction.LOG)
start_time = time.time()
sys.stderr.write(
f'\nTraining {number_of_nodes_per_layer} nodes per layer with '
f'{number_of_layers - 2 } hidden layers for '
f'{number_of_epochs} times.')
net.train(x_test, y_test, number_of_epochs)
sys.stderr.write(
f'\nTraining took {time.time() - start_time} seconds.\n'
)
test_input = [[1, 0, 0, 1, 0, 0], [1, 0, 0, 0, 1, 0],
[1, 0, 0, 0, 0, 1], [0, 1, 0, 1, 0, 0],
[0, 1, 0, 0, 1, 0], [0, 1, 0, 0, 0, 1],
[0, 0, 1, 1, 0, 0], [0, 0, 1, 0, 1, 0],
[0, 0, 1, 0, 0, 1]]
test_output = [[1, 0, 0], [0, 1, 0], [0, 0, 1], [0, 1, 0],
[0, 0, 1], [1, 0, 0], [0, 0, 1], [1, 0, 0],
[0, 1, 0]]
for test, expected in zip(test_input, test_output):
test_result = net.predict(test)
sys.stderr.write(
f'Input: {test}, Expected: {expected}, Prediction: {test_result}\n'
)
self.assertTrue(True)
class TestForMultipleHiddenLayers(unittest.TestCase):
def setUp(self):
self.patcher = patch('builtins.print')
self.mock_object = self.patcher.start()
self.x_test = np.array([[1], [0]])
self.y_test = np.array([[1], [0]])
self.hidden_node_biases = [0.10399593413625441, 0.9477095968224993]
self.output_node_biases = [0.1683069001902736]
self.weights = [
0.08946817286422626, 0.10066428298027419, 0.8768683269443115
]
self.net = NeuralNet(1,
4,
1,
1,
self.hidden_node_biases,
self.output_node_biases,
self.weights,
learning_rate=0.01)
def tearDown(self):
self.patcher.stop()
def test_weights(self):
self.assertEqual(self.net.get_weights(), self.weights)
def test_biases(self):
self.assertEqual(
self.net.get_biases(),
[0.10399593413625441, 0.9477095968224993, 0.1683069001902736])
def test_partial_derivatives_for_biases(self):
self.net.train(self.x_test, self.y_test, 1)
self.assertEqual(
self.net.get_biases(),
[0.10397901969529151, 0.9470426291307198, 0.16445586456466701])
def test_partial_derivatives_for_weights(self):
self.net.train(self.x_test, self.y_test, 1)
self.assertEqual(
self.net.get_weights(),
[0.08948139076478426, 0.10032526109304217, 0.8740529208576437])
class TestForMultipleOutputs(unittest.TestCase):
def setUp(self):
self.patcher = patch('builtins.print')
self.mock_object = self.patcher.start()
self.x_test = np.array([[1, 0], [0, 1], [0, 0], [1, 1]])
self.y_test = np.array([[0, 1], [0, 1], [1, 0], [1, 0]])
self.hidden_node_biases = [2, 2]
self.output_node_biases = [3, 3]
self.weights = [5, 5, 5, 5, 5, 5, 5, 5]
self.net = NeuralNet(2,
3,
2,
2,
self.hidden_node_biases,
self.output_node_biases,
self.weights,
activation_function=ActivationFunction.LOG,
learning_rate=0.01)
def tearDown(self):
self.patcher.stop()
def test_weights(self):
self.assertEqual(self.net.get_weights(), self.weights)
def test_biases(self):
self.assertEqual(self.net.get_biases(), [2, 2, 3, 3])
def test_partial_derivatives_for_biases(self):
self.net.train(self.x_test, self.y_test, 1)
self.assertEqual(self.net.get_biases(), [
2.000000000000002, 2.000000000000002, 3.0000000257168224,
2.99999997428318
])
def test_partial_derivatives_for_weights(self):
self.net.train(self.x_test, self.y_test, 1)
self.assertEqual(self.net.get_weights(), [
5.0, 5.0, 5.0, 5.0, 5.000000021300251, 4.999999978699751,
5.000000021300251, 4.999999978699751
])
def testForProperNumberOfWeightsInInput(self, mocked_print):
with self.assertRaises(ValueError):
NeuralNet(2, 3, 2, 1, [1, 2], [1], [1])
def testForProperNumberOfHiddenBiasesInInput(self, mocked_print):
with self.assertRaises(ValueError):
NeuralNet(2, 3, 2, 1, [1, 2, 3])
def testForProperInputType(self, mocked_print):
net = NeuralNet(2, 3, 2)
with self.assertRaises(TypeError):
net.train([[1, 0]], np.array([[1]]))
def testForProperNumberOfTrainingDataOutput(self, mocked_print):
net = NeuralNet(2, 3, 2, 3)
with self.assertRaises(ValueError):
net.check_for_valid_input(np.array([[1, 2]]), np.array([[1]]))