def test_learn_example_dataset_1(test_input):
"""Learn from the example database."""
train, _ = load_toy_cancer()
_bk = Background(modes=train.modes)
_dn = BoostedRDN(background=_bk, target="cancer", n_estimators=test_input)
_dn.fit(train)
assert len(_dn.estimators_) == test_input
def test_initialize_bad_n_estimators(test_input):
"""Test bad values for n_estimators"""
_dn = BoostedRDN(target="cancer",
background=Background(),
n_estimators=test_input)
with pytest.raises(ValueError):
_dn.fit(example_data.train)
def test_learn_example_dataset_1(test_input):
"""Learn from the example database."""
_bk = Background(modes=example_data.train.modes,
use_std_logic_variables=True)
_dn = BoostedRDN(background=_bk, target="cancer", n_estimators=test_input)
_dn.fit(example_data.train)
assert len(_dn.estimators_) == test_input
def test_predict_example_data(test_input):
"""Test learn and predict."""
train, test = load_toy_cancer()
_bk = Background(modes=train.modes)
_dn = BoostedRDN(background=_bk, target="cancer", n_estimators=test_input)
_dn.fit(train)
assert_array_equal(_dn.predict(test), np.array([1.0, 1.0, 1.0, 0.0, 0.0]))
def test_predict_example_data(test_input):
"""Test learn and predict."""
_bk = Background(modes=example_data.train.modes,
use_std_logic_variables=True)
_dn = BoostedRDN(background=_bk, target="cancer", n_estimators=test_input)
_dn.fit(example_data.train)
assert_array_equal(_dn.predict(example_data.test),
np.array([1.0, 1.0, 1.0, 0.0, 0.0]))
def test_initialize_bad_neg_pos_ratio(test_input):
"""Tests bad values for neg_pos_ratio"""
_dn = BoostedRDN(target="cancer",
background=Background(),
neg_pos_ratio=test_input)
train, _ = load_toy_cancer()
with pytest.raises(ValueError):
_dn.fit(train)
def test_toy_cancer_predict_after_load(test_input):
"""Load a ToyCancer json file and predict."""
clf = BoostedRDN()
clf.from_json(
"srlearn/tests/regression_tests/json/toy_cancer_{0}.json".format(
test_input))
_, test = load_toy_cancer()
_predictions = clf.predict(test)
assert_array_equal(_predictions, np.array([1.0, 1.0, 1.0, 0.0, 0.0]))
def test_cannot_read_outside_length_of_dotfiles():
"""Test that invalid tree indexes raise errors."""
train, _ = load_toy_cancer()
bkg = Background(modes=train.modes)
clf = BoostedRDN(target="cancer", background=bkg)
clf.fit(train)
for test_input in [-10, -5, -1, 10]:
with pytest.raises(IndexError):
_ = export_digraph(clf, tree_index=test_input)
def test_predict_proba_test_data():
"""Assert arrays are almost equal on output of predict_proba()"""
train, test = load_toy_cancer()
_bk = Background(modes=train.modes)
_dn = BoostedRDN(background=_bk, target="cancer", n_estimators=5)
_dn.fit(train)
assert_array_almost_equal(
_dn.predict_proba(test),
np.array([0.74, 0.74, 0.74, 0.25, 0.25]),
decimal=2,
)
def test_predict_proba_test_data():
"""Assert arrays are almost equal on output of predict_proba()"""
_bk = Background(modes=example_data.train.modes,
use_std_logic_variables=True)
_dn = BoostedRDN(background=_bk, target="cancer", n_estimators=5)
_dn.fit(example_data.train)
assert_array_almost_equal(
_dn.predict_proba(example_data.test),
np.array([0.74, 0.74, 0.74, 0.25, 0.25]),
decimal=2,
)
def test_feature_importances_toy_cancer():
"""Test getting the feature importances from the Toy-Cancer set."""
train, _ = load_toy_cancer()
bkg = Background(modes=train.modes)
rdn = BoostedRDN(
target="cancer",
background=bkg,
n_estimators=10,
)
rdn.fit(train)
_features = rdn.feature_importances_
assert _features.most_common(1)[0] == ("smokes", 10)
neg="datasets/imdb/train1/train1_neg.txt",
facts="datasets/imdb/train1/train1_facts.txt",
)
bk = Background(modes=[
"actor(+person).", "movie(+movie, +person).",
"movie(+movie, -person).", "movie(-movie, +person).",
"female_gender(+person).", "genre(+person, +genre).",
"genre(+person, #genre).", "genre(+person, -genre).",
"genre(-person, +genre).", "workedunder(+person, +person).",
"workedunder(+person, -person).", "workedunder(-person, +person)."
], )
clf = BoostedRDN(
background=bk,
target="workedunder",
node_size=3,
max_tree_depth=3,
)
_start = time.perf_counter()
clf.fit(db)
_end = time.perf_counter()
_difference = _end - _start
print(_difference)
_train_times.append(_difference)
print("Mean runtime for `clf.fit()` and standard deviation:")
print(np.mean(_train_times))
from srlearn.rdn import BoostedRDN
from srlearn import Background
bk = Background(
modes=[
"male(+name).", "father(+name,+name).", "childof(+name,+name).",
"siblingof(+name,+name)."
],
node_size=1,
number_of_clauses=8,
)
clf = BoostedRDN(
background=bk,
target="father",
n_estimators=5,
)
clf.fit(train)
# %%
# It's important to check whether we actually learn something useful.
# We'll visually inspect the relational regression trees to see what
# they learned.
from srlearn.plotting import plot_digraph
from srlearn.plotting import export_digraph
plot_digraph(export_digraph(clf, 0), format="html")
def test_feature_importances_before_fit():
"""Test that one cannot get feature importances before fit."""
rdn = BoostedRDN()
with pytest.raises(ValueError):
rdn.feature_importances_
def test_bad_shell_command():
"""Test running a shell command which cannot exit 0"""
_dn = BoostedRDN()
_call = "git bat"
with pytest.raises(RuntimeError):
_dn._call_shell_command(_call)
def test_initialize_bad_target(test_input):
"""Initialize an RDN with incorrect target values."""
_dn = BoostedRDN(target=test_input)
train, _ = load_toy_cancer()
with pytest.raises(ValueError):
_dn.fit(train)
def test_initialize_bad_target(test_input):
"""Initialize an RDN with incorrect target values."""
_dn = BoostedRDN(target=test_input)
with pytest.raises(ValueError):
_dn.fit(example_data.train)
def test_initialize_rdn_trees(test_input):
"""Initialize an RDN with various tree numbers."""
_dn = BoostedRDN(n_estimators=test_input)
assert _dn.n_estimators == test_input
def test_initialize_rdn_1():
"""Initialize an RDN with default parameters."""
_dn = BoostedRDN()
assert _dn.target == "None"
assert _dn.n_estimators == 10
"courseprof(+course,-person).",
"courseta(+course,-person).",
"courseta(-course,+person).",
"project(-proj,+person).",
"project(+proj,-person).",
"sameperson(-person,+person).",
"faculty(+person).",
"student(+person).",
],
number_of_clauses=8,
)
clf = BoostedRDN(
background=bkg,
target="faculty",
max_tree_depth=3,
node_size=3,
n_estimators=10,
)
clf.fit(train)
# %%
# The built-in ``feature_importances_`` attribute of a fit classifier is a
# Counter of how many times a features appears across the trees:
clf.feature_importances_
# %%
# These should generally be looked at while looking at the trees, so we'll
# plot the first tree here as well.
)
bk = Background(modes=[
"courseprof(-Course, +Person).",
"courseprof(+Course, -Person).",
"courseta(+Course, -Person).",
"courseta(-Course, +Person).",
"faculty(+Person).",
"project(-Proj, +Person).",
"project(+Proj, -Person).",
"sameperson(-Person, +Person).",
], )
clf = BoostedRDN(
background=bk,
target="faculty",
node_size=2,
max_tree_depth=3,
)
_start = time.perf_counter()
clf.fit(db)
_end = time.perf_counter()
_difference = _end - _start
print(_difference)
_train_times.append(_difference)
print("Mean runtime for `clf.fit()` and standard deviation:")
print(np.mean(_train_times))
def test_serialize_BoostedRDN(tmpdir):
"""Test that inference is possible after loading from json"""
output_json = tmpdir.join("ToyCancerRDN.json")
train, test = load_toy_cancer()
bkg = Background(modes=train.modes)
rdn = BoostedRDN(background=bkg, target="cancer", n_estimators=5)
rdn.fit(train)
rdn.to_json(output_json)
# New BoostedRDN instance, loading from file, and running.
rdn2 = BoostedRDN()
rdn2.from_json(output_json)
_predictions = rdn2.predict(test)
assert len(rdn2.estimators_) == 5
assert_array_equal(_predictions, np.array([1.0, 1.0, 1.0, 0.0, 0.0]))
def test_initialize_bad_background(test_input):
"""Test bad input for background"""
_dn = BoostedRDN(target="cancer", background=test_input)
train, _ = load_toy_cancer()
with pytest.raises(ValueError):
_dn.fit(train)
from srlearn.rdn import BoostedRDN
from srlearn import Background
from srlearn import example_data
import numpy as np
import matplotlib.pyplot as plt
bk = Background(
modes=example_data.train.modes,
use_std_logic_variables=True,
)
clf = BoostedRDN(
background=bk,
target='cancer',
max_tree_depth=2,
node_size=2,
n_estimators=20,
)
clf.fit(example_data.train)
x = np.arange(1, 21)
y_pos = []
y_neg = []
thresholds = []
for n_trees in x:
clf.set_params(n_estimators=n_trees)
probs = clf.predict_proba(example_data.test)
]
bk = Background(
modes=[
"male(+name).",
"father(+name,+name).",
"childof(+name,+name).",
"siblingof(+name,+name)."
],
number_of_clauses=8,
use_prolog_variables=True,
)
clf = BoostedRDN(
background=bk,
target="father",
n_estimators=5,
)
clf.fit(train_db)
test_db = Database()
test_db.pos = [
"father(elizabeth,mrbennet).",
"father(jane,mrbennet).",
"father(charlotte,mrlucas).",
]
test_db.neg = [
"father(charlotte,mrsbennet).",
def test_initialize_bad_background(test_input):
"""Test bad input for background"""
_dn = BoostedRDN(target="cancer", background=test_input)
with pytest.raises(ValueError):
_dn.fit(example_data.train)