def test_mlp(self):
X_train, y_train, X_test, y_test = utils.get_mnist()
model = baselines.MLP(layer_sizes=[4, 4],
input_size=X_train.shape[1],
num_classes=10)
output = model.predict(X_train[:1])
self.assertEqual(output.dtype, y_test.dtype)
self.assertEqual(output.shape, (1, ))
output = model.predict(X_train[:10], target=y_train[:10])
self.assertEqual(output.dtype, y_train.dtype)
self.assertEqual(output.shape, (10, ))
def _test_classification(self, backend):
X_train, y_train, X_test, y_test = utils.get_mnist()
model = GLN(backend=backend,
layer_sizes=[4, 4],
input_size=X_train.shape[1],
num_classes=10)
output = model.predict(X_train[:1])
self.assertTrue(np.issubdtype(output.dtype, np.integer))
self.assertEqual(output.shape, (1, ))
output = model.predict(X_train[:10], target=y_train[:10])
self.assertTrue(np.issubdtype(output.dtype, np.integer))
self.assertEqual(output.shape, (10, ))
output = model.predict(X_train[:4],
target=y_train[:4],
return_probs=True)
self.assertTrue(np.issubdtype(output.dtype, np.floating))
self.assertEqual(output.shape, (4, 10))
def _test_binary(self, backend):
X_train, y_train, X_test, y_test = utils.get_mnist()
y_train = (y_train == 0)
y_test = (y_test == 0)
model = GLN(backend=backend,
layer_sizes=[4, 4],
input_size=X_train.shape[1])
output = model.predict(X_train[:1])
self.assertEqual(output.dtype, y_test.dtype)
self.assertEqual(output.shape, (1, ))
output = model.predict(X_train[:10], target=y_train[:10])
self.assertEqual(output.dtype, y_test.dtype)
self.assertEqual(output.shape, (10, ))
output = model.predict(X_train[:4],
target=y_train[:4],
return_probs=True)
self.assertTrue(np.issubdtype(output.dtype, np.floating))
self.assertEqual(output.shape, (4, ))
def test_readme(self):
from pygln import utils
X_train, y_train, X_test, y_test = utils.get_mnist()
y_train_3 = (y_train == 3)
y_test_3 = (y_test == 3)
from pygln.numpy import GLN
model_3 = GLN(layer_sizes=[4, 4, 1],
input_size=X_train.shape[1],
learning_rate=1e-4)
# swapped
from pygln import GLN
model_3 = GLN(backend='numpy',
layer_sizes=[4, 4, 1],
input_size=X_train.shape[1],
learning_rate=1e-4)
for n in range(X_train.shape[0]):
pred = model_3.predict(X_train[n:n + 1], target=y_train_3[n:n + 1])
preds = []
batch_size = 100
for n in range(X_test.shape[0] // batch_size):
batch = X_test[n * batch_size:(n + 1) * batch_size]
pred = model_3.predict(batch)
preds.append(pred)
import numpy as np
from sklearn.metrics import accuracy_score
accuracy_score(y_test_3, np.concatenate(preds, axis=0))
model = GLN(backend='numpy',
layer_sizes=[4, 4, 1],
input_size=X_train.shape[1],
num_classes=10,
learning_rate=1e-4)
for n in range(X_train.shape[0]):
model.predict(X_train[n:n + 1], target=y_train[n:n + 1])
preds = []
for n in range(X_test.shape[0]):
preds.append(model.predict(X_test[n]))
accuracy_score(y_test, np.vstack(preds))
from pygln import utils
model_3 = GLN(backend='numpy',
layer_sizes=[4, 4, 1],
input_size=784,
learning_rate=1e-4)
print(utils.evaluate_mnist(model_3, mnist_class=3, batch_size=4))
model = GLN(backend='numpy',
layer_sizes=[4, 4, 1],
input_size=784,
num_classes=10,
learning_rate=1e-4)
print(utils.evaluate_mnist(model, batch_size=4))