def test_combine():
label_names = ['Class_1', 'Class_2', 'Class_3']
Predictions = make_multiclass(label_names=label_names)
ps_11 = [0.7, 0.1, 0.2]
ps_12 = [0.1, 0.8, 0.1]
ps_13 = [0.2, 0.5, 0.3]
predictions_1 = Predictions(y_pred=[ps_11, ps_12, ps_13])
ps_21 = [0.3, 0.4, 0.3]
ps_22 = [0.5, 0.2, 0.3]
ps_23 = [0.1, 0.5, 0.4]
predictions_2 = Predictions(y_pred=[ps_21, ps_22, ps_23])
combined = Predictions.combine(
predictions_list=[predictions_1, predictions_2], index_list=[0, 1])
ps_combined_11 = [0.5, 0.25, 0.25]
ps_combined_12 = [0.3, 0.5, 0.2]
ps_combined_13 = [0.15, 0.5, 0.35]
assert_array_almost_equal(combined.y_pred,
[ps_combined_11, ps_combined_12, ps_combined_13])
ps_31 = [-1, 0, 0.5]
ps_32 = [-1, -1, 0]
ps_33 = [0, 0.2, 0.3]
predictions_3 = Predictions(y_pred=[ps_31, ps_32, ps_33])
combined = Predictions.combine(
predictions_list=[predictions_1, predictions_3], index_list=[0, 1])
ps_combined_21 = [0.35, 0.05, 0.6]
ps_combined_22 = [0.05 + 1. / 6, 0.4 + 1. / 6, 0.05 + 1. / 6]
ps_combined_23 = [0.1, 0.45, 0.45]
assert_array_almost_equal(combined.y_pred,
[ps_combined_21, ps_combined_22, ps_combined_23])
def test_init():
label_names = ['Class_1', 'Class_2', 'Class_3']
Predictions = make_multiclass(label_names=label_names)
ps_1 = [0.7, 0.1, 0.2]
ps_2 = [0.1, 0.8, 0.1]
ps_3 = [0.2, 0.5, 0.3]
predictions = Predictions(y_pred=[ps_1, ps_2, ps_3])
assert_array_equal(predictions.y_pred, [ps_1, ps_2, ps_3])
assert_array_equal(predictions.y_pred_label_index, [0, 1, 1])
def test_multiclass_init_from_labels():
label_names = [1, 2, 3]
Predictions = make_multiclass(label_names=label_names)
y_pred_label_1 = 2
y_pred_label_2 = 1
y_pred_label_3 = 3
predictions = Predictions(
y_true=[y_pred_label_1, y_pred_label_2, y_pred_label_3])
assert_array_equal(predictions.y_pred, [[0, 1, 0], [1, 0, 0], [0, 0, 1]])
assert_array_equal(predictions.y_pred_label_index, [1, 0, 2])
def test_list_init_from_truth():
label_names = ['Class_1', 'Class_2', 'Class_3']
Predictions = make_multiclass(label_names=label_names)
y_pred_label_1 = ['Class_2']
y_pred_label_2 = ['Class_1']
y_pred_label_3 = ['Class_3']
predictions = Predictions(
y_true=[y_pred_label_1, y_pred_label_2, y_pred_label_3])
assert_array_equal(predictions.y_pred, [[0, 1, 0], [1, 0, 0], [0, 0, 1]])
assert_array_equal(predictions.y_pred_label_index, [1, 0, 2])
def test_multilabel_init_from_truth():
label_names = [1, 2, 3]
Predictions = make_multiclass(label_names=label_names)
y_pred_label_1 = [1, 3]
y_pred_label_2 = [2]
y_pred_label_3 = [1, 2, 3]
predictions = Predictions(
y_true=[y_pred_label_1, y_pred_label_2, y_pred_label_3])
assert_array_equal(predictions.y_pred,
[[0.5, 0, 0.5], [0, 1, 0], [1.0 / 3, 1.0 / 3, 1.0 / 3]])
# in case of ties, argmax returns the index of the first max
assert_array_equal(predictions.y_pred_label_index, [0, 1, 0])
def test_init_empty():
label_names = ['Class_1', 'Class_2', 'Class_3']
Predictions = make_multiclass(label_names=label_names)
predictions = Predictions(n_samples=4)
assert_array_equal(
predictions.y_pred,
np.array([
[np.nan, np.nan, np.nan],
[np.nan, np.nan, np.nan],
[np.nan, np.nan, np.nan],
[np.nan, np.nan, np.nan],
]))
def test_set_valid_in_train():
label_names = ['Class_1', 'Class_2', 'Class_3']
Predictions = make_multiclass(label_names=label_names)
predictions = Predictions(n_samples=4)
y_pred_label_1 = ['Class_2']
y_pred_label_2 = ['Class_1']
y_pred_label_3 = ['Class_3']
predictions_valid = Predictions(
y_true=[y_pred_label_1, y_pred_label_2, y_pred_label_3])
predictions.set_valid_in_train(predictions_valid, [0, 2, 3])
assert_array_equal(
predictions.y_pred,
np.array([
[0, 1, 0],
[np.nan, np.nan, np.nan],
[1, 0, 0],
[0, 0, 1],
]))
