def test_tuple_repeating_hyperparameter():
tup_hp = hyperparameter.TupleRepeatingHyperparameter(
"tuple_hyperparam",
hyperparameter.UniformIntHyperparameter("hyperparam1",
1,
5,
default=1,
n=5),
max_nrepeats=1,
default=(1, ))
expected = np.array([[(1, ), (2, ), (3, ), (4, ), (5, )]])
assert (tup_hp.get_state_space() == expected).all()
# max_nrepeats > 1
mult_tup_hp = hyperparameter.TupleRepeatingHyperparameter(
"tuple_hyperparam",
hyperparameter.UniformIntHyperparameter("hyperparam1",
1,
2,
default=1,
n=2),
max_nrepeats=2,
default=(1, ))
mult_expected = np.array([[(1, ), (2, ), (1, 1), (1, 2), (2, 1), (2, 2)]])
assert (mult_tup_hp.get_state_space() == mult_expected).all()
def test_tuple_pair_hyperparameter():
tup_hp = hyperparameter.TuplePairHyperparameter("tuple_hyperparam", [
hyperparameter.UniformIntHyperparameter(
"hyperparam1", 1, 5, default=1, n=5, as_categorical=True),
hyperparameter.UniformIntHyperparameter(
"hyperparam2", 2, 6, default=2, n=5, as_categorical=True),
],
default=(1, 2))
expected = np.array([(i, j) for i in range(1, 6) for j in range(2, 7)])
assert (tup_hp.get_state_space()["choices"] == expected).all()
def test_int_log_hyperparameter():
num_hp = hyperparameter.UniformIntHyperparameter("num_hyperparam",
1,
1000,
default=1000,
log=True,
n=4)
expected = np.array([1, 10, 100, 1000])
assert (num_hp.get_state_space() == expected).all()
def test_int_hyperparameter():
num_hp = hyperparameter.UniformIntHyperparameter("num_hyperparam",
0,
100,
default=5,
log=False,
n=5)
expected = np.array([0, 25, 50, 75, 100, 5])
assert (num_hp.get_state_space() == expected).all()
def test_int_log_hyperparameter():
num_hp = hyperparameter.UniformIntHyperparameter("num_hyperparam",
1,
1000,
default=1000,
log=True,
n=4)
state_space = num_hp.get_state_space()
assert state_space["type"] is HyperparamType.INTEGER
assert state_space["min"] == 1
assert state_space["max"] == 1000
def test_int_hyperparameter():
num_hp = hyperparameter.UniformIntHyperparameter("num_hyperparam",
0,
100,
default=5,
log=False,
n=5)
state_space = num_hp.get_state_space()
assert state_space["type"] is HyperparamType.INTEGER
assert state_space["min"] == 0
assert state_space["max"] == 100
def test_base_estimator_hyperparameter():
base_est_hp = hyperparameter.BaseEstimatorHyperparameter(
"base_estimator_hyperparam",
sklearn.tree.DecisionTreeClassifier,
hyperparameters=[
hyperparameter.UniformIntHyperparameter("max_depth",
1,
10,
default=1,
n=10,
log=False)
],
default=sklearn.tree.DecisionTreeClassifier(max_depth=1))
expected = [
sklearn.tree.DecisionTreeClassifier(max_depth=i + 1) for i in range(10)
]
for i, base_est in enumerate(base_est_hp.get_state_space()):
assert base_est.get_params() == expected[i].get_params()