def test_stacked_lstm(self):
x_train, x_test, y_train, y_test = self.data
network = algorithms.RMSProp(
[
layers.Input(self.n_time_steps),
layers.Embedding(self.n_categories, 10),
layers.LSTM(
10, only_return_final=False, weights=init.Normal(0.1)),
layers.LSTM(2, weights=init.Normal(0.1)),
layers.Sigmoid(1),
],
step=0.05,
verbose=False,
batch_size=1,
error='binary_crossentropy',
)
network.train(x_train, y_train, x_test, y_test, epochs=10)
y_predicted = network.predict(x_test).round()
accuracy = (y_predicted.T == y_test).mean()
self.assertGreaterEqual(accuracy, 0.9)
def test_lstm_modify_only_one_weight_parameter(self):
lstm_layer = layers.LSTM(
2, weights=dict(weight_in_to_ingate=init.Constant(0)))
layers.join(
layers.Input((5, 3)),
lstm_layer,
)
for key, value in lstm_layer.weights.items():
if key == 'weight_in_to_ingate':
self.assertIsInstance(value, init.Constant)
else:
self.assertIsInstance(value, init.XavierUniform)
def test_stacked_lstm_with_enabled_backwards_option(self):
x_train, x_test, y_train, y_test = self.data
x_train = x_train[:, ::-1]
x_test = x_test[:, ::-1]
network = algorithms.RMSProp(
[
layers.Input(self.n_time_steps),
layers.Embedding(self.n_categories, 10),
layers.LSTM(10, only_return_final=False, backwards=True),
layers.LSTM(2, backwards=True),
layers.Sigmoid(1),
],
step=0.1,
verbose=False,
batch_size=1,
error='binary_crossentropy',
)
network.train(x_train, y_train, x_test, y_test, epochs=20)
y_predicted = network.predict(x_test).round()
accuracy = (y_predicted.T == y_test).mean()
self.assertGreaterEqual(accuracy, 0.9)
def train_lstm(self, data, **lstm_options):
x_train, x_test, y_train, y_test = data
network = algorithms.RMSProp(
[
layers.Input(self.n_time_steps),
layers.Embedding(self.n_categories, 10),
layers.LSTM(20, **lstm_options),
layers.Sigmoid(1),
],
step=0.05,
verbose=False,
batch_size=16,
error='binary_crossentropy',
)
network.train(x_train, y_train, x_test, y_test, epochs=20)
y_predicted = network.predict(x_test).round()
accuracy = (y_predicted.T == y_test).mean()
return accuracy
def test_lstm_with_4d_input(self):
x_train, x_test, y_train, y_test = self.data
network = algorithms.RMSProp(
[
layers.Input(self.n_time_steps),
layers.Embedding(self.n_categories, 10),
# Make 4D input
layers.Reshape((self.n_time_steps, 5, 2), name='reshape'),
layers.LSTM(10),
layers.Sigmoid(1),
],
step=0.1,
verbose=False,
batch_size=1,
error='binary_crossentropy',
)
network.train(x_train, y_train, x_test, y_test, epochs=2)
reshape = network.connection.end('reshape')
# +1 for batch size
output_dimension = len(reshape.output_shape) + 1
self.assertEqual(4, output_dimension)
if __name__ == '__main__':
window_size = 40
print("Loading Shakespeare's text ...")
preprocessor = TextPreprocessing(filepath=TEXT_FILE)
n_characters = preprocessor.n_characters
x_train, x_test, y_train, y_test = preprocessor.load_samples(
window_size, stride=10)
network = algorithms.RMSProp(
[
layers.Input((window_size, n_characters)),
layers.LSTM(128),
layers.Softmax(n_characters),
],
step=0.01,
verbose=True,
batch_size=128,
loss='categorical_crossentropy',
)
network.train(x_train, y_train, x_test, y_test, epochs=10)
# Number of symbols that will be generated
n_new_symbols = 1000
# Which samples to use from the test data
test_sample_id = 0
test_sample = x_test[test_sample_id]
def test_lstm_connection_exceptions(self):
with self.assertRaises(LayerConnectionError):
layers.Input(1) > layers.LSTM(10)
def test_lstm_connection_exceptions(self):
network = layers.join(layers.LSTM(10), layers.Reshape())
with self.assertRaises(LayerConnectionError):
layers.join(layers.Input(1), network)
if __name__ == '__main__':
window_size = 40
print("Loading Shakespeare's text ...")
preprocessor = TextPreprocessing(filepath=TEXT_FILE)
n_characters = preprocessor.n_characters
x_train, x_test, y_train, y_test = preprocessor.load_samples(window_size,
stride=10)
network = algorithms.RMSProp(
[
layers.Input((window_size, n_characters)),
layers.LSTM(128, unroll_scan=True),
layers.Softmax(n_characters),
],
step=0.01,
verbose=True,
batch_size=128,
error='categorical_crossentropy',
)
network.train(x_train, y_train, x_test, y_test, epochs=10)
# Number of symbols that will be generated
n_new_symbols = 1000
# Which samples to use from the test data
test_sample_id = 0
test_sample = x_test[test_sample_id]
def test_lstm_initialization_exceptions(self):
with self.assertRaisesRegexp(ValueError, 'callable'):
layers.LSTM(1, activation_functions=dict(ingate=10))
with self.assertRaises(TypeError):
layers.LSTM(1, activation_functions=lambda x: x)
