def test_klpe_contamination():
"""Check that predict agrees with contamination parameter. """
# This requires a certain amount of data samples because the threshold is
# defined by a quantile.
X = np.random.randn(50, 2)
contamination = 0.1
clf1 = AverageKLPE(k=5, contamination=contamination)
clf1.fit(X)
assert_almost_equal(np.mean(clf1.predict(X) == 1), 1 - contamination)
clf2 = MaxKLPE(k=5, contamination=contamination)
clf2.fit(X)
assert_almost_equal(np.mean(clf2.predict(X) == 1), 1 - contamination)
def test_score_train_novelty_or_not():
"""Check score_fit_ attribute is the same if novelty=True of False"""
X = np.random.randn(50, 2)
# for AverageKLPE
clf1 = AverageKLPE(k=10)
clf2 = AverageKLPE(k=10, novelty=True)
clf1.fit(X)
clf2.fit(X)
assert_array_equal(clf1.scores_fit_, clf2.scores_fit_)
# for MaxKLPE
clf3 = MaxKLPE(k=10)
clf4 = MaxKLPE(k=10, novelty=True)
clf3.fit(X)
clf4.fit(X)
assert_array_equal(clf3.scores_fit_, clf4.scores_fit_)
def test_compute_volumes():
"""Check _compute_volumes for several masses."""
estimators = [
AverageKLPE(k=3, novelty=True),
MaxKLPE(k=3, novelty=True),
OCSVM(sigma=1.),
IsolationForest(n_estimators=5, random_state=2),
KernelSmoothing()
]
alphas = rng.randint(1, 100, size=5) / 100
alphas = np.sort(alphas)
for clf in estimators:
clf = clf.fit(X_train)
clf_test = clf.score_samples(X_test)
min_test = np.min(clf_test)
max_test = np.max(clf_test)
score_function = clf.score_samples
vols, offsets = _compute_volumes(score_function, alphas, X_test, U,
vol_tot_cube)
# check increasing order of volumes and decreasing order of offsets
assert_array_equal(vols, np.sort(vols))
assert_array_equal(offsets, -np.sort(-offsets))
# check volumes in [0, vol_tot_cube]
assert_true(np.all(0 <= vols) and np.all(vols <= vol_tot_cube))
# check offset values
assert_true(
np.all(min_test <= offsets) and np.all(offsets <= max_test))
proba_offsets_pos = (clf_test >= offsets[:, np.newaxis])
# this test requires to have a large number of samples because
# np.percentile is an empirical quantile which uses interpolation.
# this is also why we ask the values to be equal only up to the
# second decimal.
assert_array_almost_equal(np.mean(proba_offsets_pos, axis=1),
alphas,
decimal=2)
def test_maxklpe():
"""Check MaxKLPE"""
score_train_true = -np.array([np.sqrt(10), 2, np.sqrt(10)])
pred_train_true = np.array([0, 1, 0])
score_test_true = -np.array([np.sqrt(5), 2])
pred_test_true = np.array([1, 1])
# when novelty=False, i.e. scores and predict on X_train itself
clf1 = MaxKLPE(k=2, contamination=0.7)
clf1.fit(X_train)
assert_equal(clf1.algo, 'max')
score_train_attr1 = clf1.scores_fit_
assert_array_almost_equal(score_train_attr1, score_train_true)
score_train1 = clf1.score_samples(X_train)
assert_array_almost_equal(score_train1, score_train_true)
assert_array_equal((score_train1 >= clf1.threshold_).astype(int),
clf1.predict(X_train))
assert_array_equal(pred_train_true, clf1.predict(X_train))
# when novelty=True, i.e. scores and predict on X_test
clf2 = MaxKLPE(k=2, contamination=0.7, novelty=True)
clf2.fit(X_train)
score_train_attr2 = clf2.scores_fit_
assert_array_almost_equal(score_train_attr2, score_train_true)
score_test2 = clf2.score_samples(X_test)
assert_array_almost_equal(score_test2, score_test_true)
assert_array_equal(pred_test_true, clf2.predict(X_test))