def test_squeeze_raises_for_invalid_alpha(alpha: float) -> None:
kernel = gpflow.kernels.RBF(variance=1.0, lengthscales=[0.2, 0.2])
upper = tf.cast([0.5, 0.5], dtype=tf.float64)
lower = upper / 5.0
kernel.lengthscales = gpflow.Parameter(kernel.lengthscales,
transform=tfp.bijectors.Sigmoid(
low=lower, high=upper))
with pytest.raises(ValueError):
squeeze_hyperparameters(kernel, alpha)
def test_squeeze_sigmoid_hyperparameters() -> None:
kernel = gpflow.kernels.RBF(variance=1.0,
lengthscales=[0.1 + 1e-3, 0.5 - 1e-3])
upper = tf.cast([0.5, 0.5], dtype=tf.float64)
lower = upper / 5.0
kernel.lengthscales = gpflow.Parameter(kernel.lengthscales,
transform=tfp.bijectors.Sigmoid(
low=lower, high=upper))
squeeze_hyperparameters(kernel, alpha=0.1)
npt.assert_array_almost_equal(kernel.lengthscales,
[0.1 + 4e-2, 0.5 - 4e-2])
def test_squeeze_raises_for_invalid_epsilon() -> None:
lik = gpflow.likelihoods.Gaussian(variance=1.01e-6)
with pytest.raises(ValueError):
squeeze_hyperparameters(lik, epsilon=-1.0)
def test_squeeze_softplus_hyperparameters() -> None:
lik = gpflow.likelihoods.Gaussian(variance=1.01e-6)
squeeze_hyperparameters(lik, epsilon=0.2)
npt.assert_array_almost_equal(lik.variance, 0.2 + 1e-6)