def test_seed_persistence(self):
self.ae0 = AutoEncoder(seed=2345, **self.params)
self.ae1 = AutoEncoder(seed=2345, **self.params)
self.ae0.fit(X_TRAIN, epochnum=6)
self.ae1.fit(X_TRAIN, epochnum=3)
np.random.normal(size=50)
self.ae1.fit(X_TRAIN, epochnum=3)
check_ae_same_est(self.ae0, self.ae1)
class TestAEPredict:
def setup(self):
self.ae = AutoEncoder(num_inputs=3, num_comp=2, seed=323440)
self.ae.fit(X_TRAIN, epochnum=20)
@raises(AttributeError)
def test_predict_prob(self):
self.ae.predict_prob(X_TRAIN)
@raises(AttributeError)
def test_classify(self):
self.ae.classify(X_TRAIN)
def test_project(self):
project = self.ae.project(X_TRAIN)
assert project.shape == X_TRAIN.shape
def test_transform(self):
transform = self.ae.transform(X_TRAIN)
assert transform.shape[1] == 2
# %cd M:/Libraries/Documents/Code/Python/simpleml/examples
sys.path.insert(1, os.path.join(sys.path[0], '..'))
from simpleml.neural import AutoEncoder
# Load data
with gzip.open('../data/mnist.gz', 'rb') as f:
data = pickle.load(f)
# Setup autoencoder
num_comp = 25
ae = AutoEncoder(
num_inputs=data['train'][0].shape[1],
num_comp=num_comp,
learn_rate=.5, momentum=.1, seed=23456
)
# Estimate autoencoder
start = time.perf_counter()
ae.fit(data['train'][0], epochnum=1, verbose=1)
print("Time: {:5.2f}".format(time.perf_counter() - start))
# Visualize the encodings
fig1 = plt.figure(figsize=(10, 10))
for i in range(num_comp):
side = np.sqrt(num_comp)
ax = fig1.add_subplot(side, side, i+1)
ax.imshow(ae.layers[0].weights[1:, i].reshape([28, 28]), cmap=cm.Greys_r)
ax.get_xaxis().set_visible(False)
def test_same_state(self):
self.ae0 = AutoEncoder(seed=np.random.RandomState(345), **self.params)
self.ae1 = AutoEncoder(seed=np.random.RandomState(345), **self.params)
self.ae0.fit(X_TRAIN, epochnum=3)
self.ae1.fit(X_TRAIN, epochnum=3)
check_ae_same_est(self.ae0, self.ae1)
def test_same_seed(self):
self.ae0 = AutoEncoder(seed=2345, **self.params)
self.ae1 = AutoEncoder(seed=2345, **self.params)
self.ae0.fit(X_TRAIN, epochnum=3)
self.ae1.fit(X_TRAIN, epochnum=3)
check_ae_same_est(self.ae0, self.ae1)
class TestAESeed:
params = {'num_inputs': 3, 'num_comp': 2}
def test_same_seed(self):
self.ae0 = AutoEncoder(seed=2345, **self.params)
self.ae1 = AutoEncoder(seed=2345, **self.params)
self.ae0.fit(X_TRAIN, epochnum=3)
self.ae1.fit(X_TRAIN, epochnum=3)
check_ae_same_est(self.ae0, self.ae1)
def test_same_state(self):
self.ae0 = AutoEncoder(seed=np.random.RandomState(345), **self.params)
self.ae1 = AutoEncoder(seed=np.random.RandomState(345), **self.params)
self.ae0.fit(X_TRAIN, epochnum=3)
self.ae1.fit(X_TRAIN, epochnum=3)
check_ae_same_est(self.ae0, self.ae1)
def test_seed_persistence(self):
self.ae0 = AutoEncoder(seed=2345, **self.params)
self.ae1 = AutoEncoder(seed=2345, **self.params)
self.ae0.fit(X_TRAIN, epochnum=6)
self.ae1.fit(X_TRAIN, epochnum=3)
np.random.normal(size=50)
self.ae1.fit(X_TRAIN, epochnum=3)
check_ae_same_est(self.ae0, self.ae1)
def setup(self):
self.ae = AutoEncoder(num_inputs=3, num_comp=2, seed=323440)
self.ae.fit(X_TRAIN, epochnum=20)
def test_fit(self):
x = np.array([[0.1, 0.1], [0, 0.1], [1.1, 0.9], [1, 1.1]])
ae = AutoEncoder(num_inputs=2, num_comp=1, seed=323440)
ae.fit(x, epochnum=20)
def test_nodes(self):
ae = AutoEncoder(num_inputs=5, num_comp=3)
assert ae.layers[0].num_inputs == 6
assert ae.layers[0].num_nodes == 4
