def test_compute_cost():
y = np.asarray([[1, 1, 1]])
al = np.array([[.8, .9, 0.4]])
nn = NeuralNetwork(layer_dims=(3, 1), params_ok=True)
cost = nn.compute_cost(al, y)
assert_eq(cost, 0.41493159961539694)
def test_model_backward():
np.random.seed(3)
al = np.random.randn(1, 2)
y = np.array([[1, 0]])
a1 = np.random.randn(4, 2)
w1 = np.random.randn(3, 4)
b1 = np.random.randn(3, 1)
z1 = np.random.randn(3, 2)
a2 = np.random.randn(3, 2)
w2 = np.random.randn(1, 3)
b2 = np.random.randn(1, 1)
z2 = np.random.randn(1, 2)
nn = NeuralNetwork(w=[w1, w2], b=[b1, b2], params_ok=True)
nn.cache = [(a1, z1), (a2, z2)]
nn.w = [w1, w2]
nn.b = [b1, b2]
grads = nn.model_backward(al, y)
assert_eq(grads[0][0], [
[0.41010002, 0.07807203, 0.1379844364, 0.1050216745],
[0., 0., 0., 0.],
[0.05283652, 0.0100586544, 0.017777656, 0.0135307956]
])
assert_eq(grads[1][0], [
[-0.2200706339], [0.], [-0.02835349]
])
def test_compute_cost():
Y = np.asarray([[1, 1, 1]])
A = np.asarray([[.8, .9, .4]])
nn = NeuralNetwork(lostfunc='binary_crossentropy')
cost = nn.compute_cost(A, Y)
assert_eq(cost, 0.41493160)
def test_point_point():
p1 = Point(0,0)
p2 = Point(1,1)
ok_(p1.geom_type=='Point', p1.geom_type)
assert_ae(p1.coords[:][0], (0,0))
d = p1.distance(p2)
assert_eq(d, sqrt(2))
def test_initialize_he():
np.random.seed(3)
nn = NeuralNetwork(layer_dims=[2, 4, 1], weight_factor='he', params_ok=True)
assert_eq(nn.w[0], [
[1.78862847, 0.43650985],
[0.09649747, -1.8634927],
[-0.2773882, -0.35475898],
[-0.08274148, -0.62700068]
], rtol=1e-5)
assert_eq(nn.w[1], [
[-0.03098412, -0.33744411, -0.92904268, 0.62552248]
], rtol=1e-5)
def test_linear_activate_forward():
np.random.seed(2)
a_prev = np.random.randn(3, 2)
w = np.random.randn(1, 3)
b = np.random.randn(1, 1)
nn = NeuralNetwork(w=[w], b=[b], params_ok=True)
a = nn.forward(a_prev, w, b, nn.activation_funcs['sigmoid'][0])
assert_eq(a, [[0.96890023, 0.11013289]])
nn = NeuralNetwork(w=[w], b=[b], params_ok=True)
a = nn.forward(a_prev, w, b, nn.activation_funcs['relu'][0])
assert_eq(a, [[3.43896131, 0.]])
def test_model_forward():
np.random.seed(6)
x = np.random.randn(5, 4)
w1 = np.random.randn(4, 5)
b1 = np.random.randn(4, 1)
w2 = np.random.randn(3, 4)
b2 = np.random.randn(3, 1)
w3 = np.random.randn(1, 3)
b3 = np.random.randn(1, 1)
nn = NeuralNetwork(w=[w1, w2, w3], b=[b1, b2, b3], params_ok=True)
al = nn.model_forward(x)
nt.eq_(len(nn.cache), 3)
assert_eq(al, [[0.03921668, 0.70498921, 0.19734387, 0.04728177]])
def test_layer_forward():
np.random.seed(2)
Aprev = np.random.randn(3, 2)
W = np.random.randn(1, 3)
b = np.random.randn(1, 1)
layer = DenseLayer(W, b)
A = layer.forward(Aprev)
assert_eq(A, [[3.43896131, 0.]])
layer = DenseLayer(W, b, activation='sigmoid')
A = layer.forward(Aprev)
assert_eq(A, [[0.96890023, 0.11013289]])
def test_update_parameters():
np.random.seed(2)
w1 = np.random.randn(3, 4)
b1 = np.random.randn(3, 1)
w2 = np.random.randn(1, 3)
b2 = np.random.randn(1, 1)
np.random.seed(3)
dw1 = np.random.randn(3, 4)
db1 = np.random.randn(3, 1)
dw2 = np.random.randn(1, 3)
db2 = np.random.randn(1, 1)
nn = NeuralNetwork(w=[w1, w2], b=[b1, b2], learning_rate=0.1, params_ok=True)
nn.update_parameters(([dw1, dw2], [db1, db2]))
assert_eq(nn.w[0], [
[-0.5956206947, -0.09991781, -2.14584584, 1.82662008],
[-1.7656967649, -0.80627147, 0.5111555653, -1.18258802],
[-1.0535704, -0.86128581, 0.68284052, 2.2037457748]
])
assert_eq(nn.w[1], [
[-0.55569196, 0.0354055, 1.32964895]
])
assert_eq(nn.b[0], [
[-0.04659241], [-1.28888275], [0.53405496]
])
assert_eq(nn.b[1], [[-0.84610769]])
def test_shaped_instance(self, seed):
"""Use shape keyword argument."""
dim = Dimension('yolo', 'norm', 0.9, shape=(3, 2))
samples = dim.sample(seed=seed)
assert len(samples) == 1
assert_eq(dists.norm.rvs(0.9, size=(3, 2)), samples[0])
assert dim.shape == (3, 2)
dim = Dimension('yolo', 'norm', 0.9, shape=4)
samples = dim.sample(seed=seed)
assert len(samples) == 1
assert_eq(dists.norm.rvs(0.9, size=4), samples[0])
assert dim.shape == (4, )
def test_forward_propagation_dropout():
np.random.seed(1)
x = np.random.randn(3, 5)
w1 = np.random.randn(2, 3)
b1 = np.random.randn(2, 1)
w2 = np.random.randn(3, 2)
b2 = np.random.randn(3, 1)
w3 = np.random.randn(1, 3)
b3 = np.random.randn(1, 1)
np.random.seed(1)
nn = NeuralNetwork(w=[w1, w2, w3], b=[b1, b2, b3], params_ok=True)
a3 = nn.model_forward(x, keep_prop=0.7)
assert_eq(a3,
[[0.36974721, 0.00305176, 0.04565099, 0.49683389, 0.36974721]],
rtol=1e-5)
def test_cost_regulation():
np.random.seed(1)
y = np.array([[1, 1, 0, 1, 0]])
w1 = np.random.randn(2, 3)
b1 = np.random.randn(2, 1)
w2 = np.random.randn(3, 2)
b2 = np.random.randn(3, 1)
w3 = np.random.randn(1, 3)
b3 = np.random.randn(1, 1)
a3 = np.array(
[[0.40682402, 0.01629284, 0.16722898, 0.10118111, 0.40682402]])
nn = NeuralNetwork(w=[w1, w2, w3],
b=[b1, b2, b3],
lambd=0.1,
params_ok=True)
cost = nn.compute_cost(a3, y)
assert_eq(cost, 1.78648594516)
def test_forward():
np.random.seed(6)
nn = NeuralNetwork(lostfunc='binary_crossentropy')
X = np.random.randn(5, 4)
W = np.random.randn(4, 5)
b = np.random.randn(4, 1)
nn.add_layer(DenseLayer(W, b))
W = np.random.randn(3, 4)
b = np.random.randn(3, 1)
nn.add_layer(DenseLayer(W, b))
W = np.random.randn(1, 3)
b = np.random.randn(1, 1)
nn.add_layer(DenseLayer(W, b, activation='sigmoid'))
A = nn.forward(X)
assert_eq(A, [[0.03921668, 0.70498921, 0.19734387, 0.04728177]])
def test_backward_propagation_dropout():
np.random.seed(1)
x = np.random.randn(3, 5)
y = np.array([[1, 1, 0, 1, 0]])
w1 = np.array([[-1.09989127, -0.17242821, -0.87785842],
[0.04221375, 0.58281521, -1.10061918]])
b1 = np.array([[1.14472371], [0.90159072]])
w2 = np.array([[0.50249434, 0.90085595], [-0.68372786, -0.12289023],
[-0.93576943, -0.26788808]])
b2 = np.array([[0.53035547], [-0.69166075], [-0.39675353]])
w3 = np.array([[-0.6871727, -0.84520564, -0.67124613]])
b3 = np.array([[-0.0126646]])
nn = NeuralNetwork(w=[w1, w2, w3], b=[b1, b2, b3], params_ok=True)
z1 = np.array(
[[-1.52855314, 3.32524635, 2.13994541, 2.60700654, -0.75942115],
[-1.98043538, 4.1600994, 0.79051021, 1.46493512, -0.45506242]])
d1 = np.array(
[[True, False, True, True, True], [True, True, True, True, False]],
dtype=bool)
a1 = np.array([[0., 0., 4.27989081, 5.21401307, 0.],
[0., 8.32019881, 1.58102041, 2.92987024, 0.]])
z2 = np.array(
[[0.53035547, 8.02565606, 4.10524802, 5.78975856, 0.53035547],
[-0.69166075, -1.71413186, -3.81223329, -4.61667916, -0.69166075],
[-0.39675353, -2.62563561, -4.82528105, -6.0607449, -0.39675353]])
d2 = np.array(
[[True, False, True, False, True], [False, True, False, True, True],
[False, False, True, False, False]],
dtype=bool)
a2 = np.array([
[1.06071093, 0., 8.21049603, 0., 1.06071093],
[0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0.],
])
z3 = np.array(
[[-0.7415562, -0.0126646, -5.65469333, -0.0126646, -0.7415562]])
a3 = np.array(
[[0.32266394, 0.49683389, 0.00348883, 0.49683389, 0.32266394]])
nn.cache = [(x, z1, d1), (a1, z2, d2), (a2, z3)]
gradw, _ = nn.model_backward(a3, y, keep_prop=0.8)
assert_eq(gradw[0], [[0.0001988393, 0.0002865694, 0.0001213795],
[0.0003564729, 0.0005137526, 0.0002176053]],
rtol=1e-5)
assert_eq(gradw[1], [[-0.0025651848, -0.0009475965], [0., 0.], [0., 0.]],
rtol=1e-5)
assert_eq(gradw[2], [[-0.0695119105, 0., 0.]])
def test_backward_propagation_regulation():
np.random.seed(1)
x = np.random.randn(3, 5)
y = np.array([[1, 1, 0, 1, 0]])
w1 = np.array([[-1.09989127, -0.17242821, -0.87785842],
[0.04221375, 0.58281521, -1.10061918]])
b1 = np.array([[1.14472371], [0.90159072]])
w2 = np.array([[0.50249434, 0.90085595], [-0.68372786, -0.12289023],
[-0.93576943, -0.26788808]])
b2 = np.array([[0.53035547], [-0.69166075], [-0.39675353]])
w3 = np.array([[-0.6871727, -0.84520564, -0.67124613]])
b3 = np.array([[-0.0126646]])
z1 = np.array(
[[-1.52855314, 3.32524635, 2.13994541, 2.60700654, -0.75942115],
[-1.98043538, 4.1600994, 0.79051021, 1.46493512, -0.45506242]])
a1 = np.array([[0., 3.32524635, 2.13994541, 2.60700654, 0.],
[0., 4.1600994, 0.79051021, 1.46493512, 0.]])
z2 = np.array(
[[0.53035547, 5.94892323, 2.31780174, 3.16005701, 0.53035547],
[-0.69166075, -3.47645987, -2.25194702, -2.65416996, -0.69166075],
[-0.39675353, -4.62285846, -2.61101729, -3.22874921, -0.39675353]])
a2 = np.array(
[[0.53035547, 5.94892323, 2.31780174, 3.16005701, 0.53035547],
[0., 0., 0., 0., 0.], [0., 0., 0., 0., 0.]])
z3 = np.array(
[[-0.3771104, -4.10060224, -1.60539468, -2.18416951, -0.3771104]])
a3 = np.array(
[[0.40682402, 0.01629284, 0.16722898, 0.10118111, 0.40682402]])
nn = NeuralNetwork(w=[w1, w2, w3],
b=[b1, b2, b3],
lambd=0.7,
params_ok=True)
nn.cache = [(x, z1), (a1, z2), (a2, z3)]
gradw, gradb = nn.model_backward(a3, y)
assert_eq(gradw[0], [[-0.256046467, 0.122988299, -0.28297132],
[-0.17706304, 0.34536100, -0.4410572]],
rtol=1e-5)
assert_eq(gradw[1], [[0.792764876, 0.85133918], [-0.0957219, -0.01720463],
[-0.13100772, -0.03750433]],
rtol=1e-5)
assert_eq(gradw[2], [[-1.77691347, -0.11832879, -0.09397446]], rtol=1e-5)
def test_layer_backward():
np.random.seed(2)
dA = np.random.randn(1, 2)
A = np.random.randn(3, 2)
W = np.random.randn(1, 3)
b = np.random.randn(1, 1)
Z = np.random.randn(1, 2)
layer = DenseLayer(np.array(W), np.array(b))
layer.Aprev = A
layer.Z = Z
dAprev = layer.backward(dA, learning_rate=0.0075)
assert_eq(dAprev, [
[0.44090989, 0.],
[0.37883606, 0.],
[-0.2298228, -0.]
])
assert_eq(layer.W, [[-1.06129076, -0.91181047, 0.55223997]])
assert_eq(layer.b, [[2.29377085]])
layer = DenseLayer(np.array(W), np.array(b), activation='sigmoid')
layer.Aprev = A
layer.Z = Z
dAprev = layer.backward(dA, learning_rate=0.0075)
assert_eq(dAprev, [
[0.110179936, 0.011053395],
[0.094668170, 0.009497234],
[-0.057430922, -0.005761545]
])
assert_eq(layer.W, [[-1.05872223, -0.90974101, 0.55160165]])
assert_eq(layer.b, [[2.29263773]])
def test_many_samples(self, seed):
"""More than 1."""
dim = Dimension('yolo', 'uniform', -3, 4, shape=(4, 4))
samples = dim.sample(n_samples=4, seed=seed)
assert len(samples) == 4
assert_eq(dists.uniform.rvs(-3, 4, size=(4, 4)), samples[0])
def _assert_target(x, target):
assert_eq(x.dims, CONF["data_dims"][target])
assert_string_equal(x.name, target)
def test_linear_activate_backward():
np.random.seed(2)
da = np.random.randn(1, 2)
a = np.random.randn(3, 2)
w = np.random.randn(1, 3)
b = np.random.randn(1, 1)
z = np.random.randn(1, 2)
nn = NeuralNetwork(w=[w], b=[b], params_ok=True)
da_prev, dw, db = nn.backward(da, w, b, (a, z), nn.activation_funcs['sigmoid'][1])
assert_eq(da_prev, [
[0.110179936, 0.011053395],
[0.094668170, 0.009497234],
[-0.057430922, -0.005761545]
])
assert_eq(dw, [[0.102667864, 0.09778551, -0.019680842]])
nt.eq_(db.shape, (1, 1))
assert_eq(db, -0.057296222)
nn = NeuralNetwork(w=[w], b=[b], params_ok=True)
da_prev, dw, db = nn.backward(da, w, b, (a, z), nn.activation_funcs['relu'][1])
assert_eq(da_prev, [
[0.44090989, 0.],
[0.37883606, 0.],
[-0.2298228, -0.]
])
assert_eq(dw, [[0.44513824, 0.37371418, -0.10478989]])
nt.eq_(db.shape, (1, 1))
assert_eq(db, -0.20837892)
def test_initialize():
np.random.seed(1)
nn = NeuralNetwork(layer_dims=[3, 2, 1], params_ok=True)
assert_eq(nn.w[0], [
[0.01624345, -0.00611756, -0.00528172],
[-0.01072969, 0.00865408, -0.02301539]
], rtol=1e-6)
assert_eq(nn.b[0], [[0], [0]])
assert_eq(nn.w[1], [[0.01744812, -0.00761207]], rtol=1e-6)
nt.eq_(nn.b[1].shape, (1, 1))
assert_eq(nn.b[1], [[0.]])
np.random.seed(3)
nn = NeuralNetwork(layer_dims=[5, 4, 3], params_ok=True)
assert_eq(nn.w[0], [
[0.01788628, 0.0043651, 0.00096497, -0.01863493, -0.00277388],
[-0.00354759, -0.00082741, -0.00627001, -0.00043818, -0.00477218],
[-0.01313865, 0.00884622, 0.00881318, 0.01709573, 0.00050034],
[-0.00404677, -0.0054536, -0.01546477, 0.00982367, -0.01101068]
], rtol=1e-5)
assert_eq(nn.b[0], [[0.], [0.], [0.], [0.]])
assert_eq(nn.w[1], [
[-0.01185047, -0.0020565, 0.01486148, 0.00236716],
[-0.01023785, -0.00712993, 0.00625245, -0.00160513],
[-0.00768836, -0.00230031, 0.00745056, 0.01976111]
], rtol=1e-5)
assert_eq(nn.b[1], [[0.], [0.], [0.]])
