def test_lipschitz_constant_loss_mse():
rng = check_random_state(42)
X, _, w, mask = _make_data(rng=rng, masked=True)
l1_ratio = 1.
alpha = .1
mask = np.ones(X.shape[1]).astype(bool)
grad_weight = alpha * X.shape[0] * (1. - l1_ratio)
a = _squared_loss_derivative_lipschitz_constant(X, mask, grad_weight)
b = spectral_norm_squared(X)
np.testing.assert_almost_equal(a, b)
def test_lipschitz_constant_loss_mse():
rng = check_random_state(42)
X, _, w, mask = _make_data(rng=rng, masked=True)
l1_ratio = 1.
alpha = .1
mask = np.ones(X.shape[1]).astype(np.bool)
grad_weight = alpha * X.shape[0] * (1. - l1_ratio)
a = _squared_loss_derivative_lipschitz_constant(X, mask, grad_weight)
b = spectral_norm_squared(X)
np.testing.assert_almost_equal(a, b)
def test_squared_loss_lipschitz(n_samples=4, n_features=2, random_state=42):
rng = np.random.RandomState(random_state)
for scaling in np.logspace(-3, 3, num=7):
X = rng.randn(n_samples, n_features) * scaling
y = rng.randn(n_samples)
n_features = X.shape[1]
L = spectral_norm_squared(X)
_check_lipschitz_continuous(lambda w: _squared_loss_grad(X, y, w),
n_features, L)
def test_squared_loss_lipschitz(n_samples=4, n_features=2, random_state=42):
rng = np.random.RandomState(random_state)
for scaling in np.logspace(-3, 3, num=7):
X = rng.randn(n_samples, n_features) * scaling
y = rng.randn(n_samples)
n_features = X.shape[1]
L = spectral_norm_squared(X)
_check_lipschitz_continuous(lambda w: _squared_loss_grad(
X, y, w), n_features, L)
