def test_backward_pass():
npr.seed(1)
N = 10
D = 3
alpha = Hyperparameter(
initial_value = 2*np.ones(D),
prior = priors.Lognormal(1.5),
name = 'alpha'
)
beta = Hyperparameter(
initial_value = 0.5*np.ones(D),
prior = priors.Lognormal(1.5),
name = 'beta'
)
bw = KumarWarp(D, alpha=alpha, beta=beta)
data = 0.5*np.ones(D)
v = npr.randn(D)
bw.forward_pass(data)
assert np.all(bw.backward_pass(v) == 0.5773502691896257*v)
def test_backward_pass():
npr.seed(1)
N = 10
D = 3
alpha = Hyperparameter(
initial_value = 2*np.ones(D),
prior = priors.Lognormal(1.5),
name = 'alpha'
)
beta = Hyperparameter(
initial_value = 0.5*np.ones(D),
prior = priors.Lognormal(1.5),
name = 'beta'
)
bw = KumarWarp(D, alpha=alpha, beta=beta)
data = 0.5*np.ones(D)
v = npr.randn(D)
bw.forward_pass(data)
assert np.all(bw.backward_pass(v) == 0.5773502691896257*v)
def test_forward_pass():
npr.seed(1)
N = 10
D = 3
alpha = Hyperparameter(initial_value=2 * np.ones(D),
prior=priors.Lognormal(1.5),
name='alpha')
beta = Hyperparameter(initial_value=0.5 * np.ones(D),
prior=priors.Lognormal(1.5),
name='beta')
bw = KumarWarp(D, alpha=alpha, beta=beta)
data = 0.5 * np.ones(D)
assert np.all(bw.forward_pass(data) == 0.1339745962155614)
def test_validation():
warnings.filterwarnings('error')
npr.seed(1)
N = 10
D = 3
kw = KumarWarp(D)
data = npr.randn(N,D)
assert_raises(UserWarning, kw.forward_pass, data)
def test_forward_pass():
npr.seed(1)
N = 10
D = 3
alpha = Hyperparameter(
initial_value = 2*np.ones(D),
prior = priors.Lognormal(1.5),
name = 'alpha'
)
beta = Hyperparameter(
initial_value = 0.5*np.ones(D),
prior = priors.Lognormal(1.5),
name = 'beta'
)
bw = KumarWarp(D, alpha=alpha, beta=beta)
data = 0.5*np.ones(D)
assert np.all(bw.forward_pass(data) == 0.1339745962155614)
