def test_score_to_label(self):
manual_scores = [0.1, 0.4, 0.2, 0.3, 0.5, 0.9, 0.7, 1, 0.8, 0.6]
labels = score_to_label(manual_scores, outliers_fraction=0.1)
assert_allclose(labels, [0, 0, 0, 0, 0, 0, 0, 1, 0, 0])
labels = score_to_label(manual_scores, outliers_fraction=0.3)
assert_allclose(labels, [0, 0, 0, 0, 0, 1, 0, 1, 1, 0])
def _parallel_predict(n_estimators, clfs, approximators, X, total_n_estimators,
rp_transformers, approx_flags, contamination, verbose):
X = check_array(X)
pred = []
for i in range(n_estimators):
estimator = clfs[i]
if verbose > 1:
print("predicting with estimator %d of %d for this parallel run "
"(total %d)..." % (i + 1, n_estimators, total_n_estimators))
# project matrix
X_scaled = jl_transform(X, rp_transformers[i])
# turn approximator scores to labels by outlier
if approx_flags[i] == 1:
predicted_labels = score_to_label(
approximators[i].predict(X_scaled),
outliers_fraction=contamination)
else:
predicted_labels = estimator.predict(X_scaled)
pred.append(predicted_labels)
return pred
def _parallel_predict(n_estimators, clfs, approximators, X, total_n_estimators,
rp_flags, rp_transformers, approx_flags, contamination,
verbose):
X = check_array(X)
pred = []
for i in range(n_estimators):
estimator = clfs[i]
if verbose > 1:
print("predicting with estimator %d of %d for this parallel run "
"(total %d)..." % (i + 1, n_estimators, total_n_estimators))
# if the random projection is needed
if rp_flags[i] == 1:
X_scaled = jl_transform(X, rp_transformers[i])
else:
X_scaled = X
# turn approximator scores to labels by outlier
# todo: decide whether the approximation should happen on the reduced
# space or the original space. For now, it is on the original space
if approx_flags[i] == 1:
predicted_labels = score_to_label(approximators[i].predict(X),
outliers_fraction=contamination)
else:
predicted_labels = estimator.predict(X_scaled)
pred.append(predicted_labels)
return pred