def test_awn_fit():
X = np.random.standard_normal((10, 2))
Z = np.random.standard_normal((10, 1))
loss = lambda target, prediction: squared(target, prediction[:, :target.shape[1]])
mlp = AwnNetwork(
2, [10], 1, ['rectifier'], 'identity', loss, max_iter=10)
mlp.fit(X, Z)
def test_linear_regression():
inpt = T.matrix('inpt')
inpt.tag.test_value = np.zeros((3, 10))
inpt.tag.test_value
target = T.matrix('target')
target.tag.test_value = np.zeros((3, 2))
l = AffineNonlinear(inpt, 10, 2, 'tanh')
loss = squared(target, l.output).sum(1).mean()
m = SupervisedModel(inpt=inpt, target=target, output=l.output, loss=loss,
parameters=l.parameters)
f_predict = m.function([m.inpt], m.output)
f_loss = m.function([m.inpt, m.target], m.loss)
X = np.zeros((20, 10))
Z = np.zeros((20, 2))
Y = f_predict(X)
assert Y.shape == (20, 2), 'ouput has wrong shape'
l = f_loss(X, Z)
assert np.array(l).ndim == 0, 'loss is not a scalar'
def test_fd_predict():
X = np.random.standard_normal((10, 2))
X, = theano_floatx(X)
loss = lambda target, prediction: squared(target, prediction[:, :target.shape[1]])
mlp = FastDropoutNetwork(
2, [10], 1, ['rectifier'], 'identity', loss, max_iter=10)
mlp.predict(X)
def test_linear_regression():
inpt = T.matrix('inpt')
inpt.tag.test_value = np.zeros((3, 10))
inpt.tag.test_value
target = T.matrix('target')
target.tag.test_value = np.zeros((3, 2))
l = AffineNonlinear(inpt, 10, 2, 'tanh')
loss = squared(target, l.output).sum(1).mean()
m = SupervisedModel(inpt=inpt,
target=target,
output=l.output,
loss=loss,
parameters=l.parameters)
f_predict = m.function([m.inpt], m.output)
f_loss = m.function([m.inpt, m.target], m.loss)
X = np.zeros((20, 10))
Z = np.zeros((20, 2))
Y = f_predict(X)
assert Y.shape == (20, 2), 'ouput has wrong shape'
l = f_loss(X, Z)
assert np.array(l).ndim == 0, 'loss is not a scalar'
def test_squared():
X, Y = T.matrix(), T.matrix()
X.tag.test_value = test_X
Y.tag.test_value = test_Y
dist = squared(X, Y).sum()
f = theano.function([X, Y], dist, mode='FAST_COMPILE')
res = f(test_X, test_Y)
assert np.allclose(res, 0.0097), 'squared loss not working'
def test_fd_iter_fit():
X = np.random.standard_normal((10, 2))
Z = np.random.standard_normal((10, 1))
loss = lambda target, prediction: squared(target, prediction[:, :target.shape[1]])
mlp = FastDropoutNetwork(
2, [10], 1, ['rectifier'], 'identity', loss, max_iter=10)
for i, info in enumerate(mlp.iter_fit(X, Z)):
if i >= 10:
break
def test_squared_colwise():
X, Y = T.matrix(), T.matrix()
X.tag.test_value = test_X
Y.tag.test_value = test_Y
dist = squared(X, Y).sum(axis=0)
f = theano.function([X, Y], dist, mode='FAST_COMPILE')
res = f(test_X, test_Y)
correct = np.allclose(res, [0.0018, 0.005, 0.0029])
assert correct, 'squared loss colwise not working'
def test_squared_rowwise():
X, Y = T.matrix(), T.matrix()
X.tag.test_value = test_X
Y.tag.test_value = test_Y
dist = squared(X, Y).sum(axis=1)
f = theano.function([X, Y], dist, mode='FAST_COMPILE')
res = f(test_X, test_Y)
correct = roughly(res, [0.0014, 0.0083])
assert correct, 'squared loss rowwise not working'
def test_fd_iter_fit():
X = np.random.standard_normal((10, 2))
Z = np.random.standard_normal((10, 1))
X, Z = theano_floatx(X, Z)
loss = lambda target, prediction: squared(target, prediction[:, :target.shape[1]])
mlp = FastDropoutNetwork(
2, [10], 1, ['rectifier'], 'identity', loss, max_iter=10)
for i, info in enumerate(mlp.iter_fit(X, Z)):
if i >= 10:
break