def test_can_set_scipy_distribution():
p = Identity().set_hyperparams_space(HyperparameterSpace({
'rand_int_scipy': randint(low=2, high=5), # scipy
'rand_int_neuraxle': RandInt(2, 5), # neuraxle
'gamma_scipy': gamma(0.2), # scipy
}))
assert isinstance(p.get_hyperparams_space()['rand_int_scipy'], ScipyDiscreteDistributionWrapper)
assert isinstance(p.get_hyperparams_space()['gamma_scipy'], ScipyContinuousDistributionWrapper)
randint_sample = p.get_hyperparams_space()['rand_int_scipy'].rvs()
gamma_sample = p.get_hyperparams_space()['gamma_scipy'].rvs()
assert 5 >= randint_sample >= 2
assert isinstance(gamma_sample, float)
def test_randint():
hd = RandInt(min_included=-10, max_included=10, null_default_value=0)
_test_randint(hd)
assert hd.max() == len(choice_list) - 1
assert abs(hd.mean() - (len(choice_list) - 1) / 2) < 1e-6
assert abs(hd.var() - (len(choice_list)**2 - 1) / 12) < 1e-6
assert abs(hd.std() - math.sqrt((len(choice_list)**2 - 1) / 12)) < 1e-6
# Convert samples in sample index
samples_index = [
get_index_in_list_with_bool(choice_list, sample) for sample in samples
]
# Verify that hd mean and variance also correspond to mean and variance of sampling.
assert abs((hd.mean() - np.mean(samples_index)) / hd.mean()) < 1e-1
assert abs((hd.var() - np.var(samples_index)) / hd.var()) < 1e-1
@pytest.mark.parametrize(
"hd, test_method",
[(RandInt(min_included=-10, max_included=10,
null_default_value=0), _test_randint),
(LogNormal(hard_clip_min=-5,
hard_clip_max=5,
log2_space_mean=0.0,
log2_space_std=2.0,
null_default_value=-1.0), _test_lognormal),
(Normal(hard_clip_min=0.0,
hard_clip_max=1.0,
mean=0.5,
std=0.2,
null_default_value=0.0), _test_normal),
(LogUniform(min_included=0.001, max_included=10), _test_loguniform),
(Uniform(min_included=-10, max_included=10), _test_uniform),
(Poisson(
min_included=0.0, max_included=10.0, null_default_value=0.0,
mu=5.0), _test_discrete_poisson),