def test_non_consecutive_labels_ari():
"""regression tests for labels with gaps
"""
ari_1 = adjusted_rand_score([0, 0, 0, 1, 1, 1], [0, 1, 0, 1, 2, 2])
ari_2 = adjusted_rand_score([0, 0, 0, 1, 1, 1], [0, 4, 0, 4, 2, 2])
assert_almost_equal(ari_1, 0.24, 2)
assert_almost_equal(ari_2, 0.24, 2)
def test_RxC_metrics():
"""Alternative implementations should coincide for RxC matrices
"""
for _ in xrange(100):
ltrue = np.random.randint(low=0, high=5, size=(20,))
lpred = np.random.randint(low=0, high=5, size=(20,))
cm = ClusteringMetrics.from_labels(ltrue, lpred)
# homogeneity, completeness, V-measure
expected_v = cm.vi_similarity_m3()
expected_hcv = sklearn_hcv(ltrue, lpred)
actual_hcv = cm.entropy_scores()
assert_array_almost_equal(actual_hcv, expected_hcv)
assert_array_almost_equal(actual_hcv[2], expected_v)
# mutual information score
expected_mi = sklearn_mi(ltrue, lpred)
actual_mi = mutual_info_score(ltrue, lpred)
assert_array_almost_equal(actual_mi, expected_mi)
# adjusted mutual information
expected_ami = sklearn_ami(ltrue, lpred)
actual_ami = adjusted_mutual_info_score(ltrue, lpred)
assert_array_almost_equal(actual_ami, expected_ami)
# adjusted rand index
expected_ari = sklearn_ari(ltrue, lpred)
actual_ari = adjusted_rand_score(ltrue, lpred)
assert_array_almost_equal(actual_ari, expected_ari)
def test_ari_nan():
"""Returns NaN for empty lists
"""
ari = adjusted_rand_score([], [])
assert_true(np.isnan(ari))
