fix tests

tests/test_intersection_based/test_intersection_based_intermediate_statistics.py

@@ -160,45 +160,6 @@ def test_accumulated_statistics(t_step, num_jobs):
     ), (stats['cts']['b'], expected_cross_triggers)
 
 
-def test_no_ground_truth_events():
-    detection_scores = np.concatenate((np.arange(5), np.arange(4)[::-1]))
-    timestamps = np.arange(10)/10
-    scores = pd.DataFrame(
-        np.array((
-            timestamps[:-1], timestamps[1:],
-            detection_scores, np.zeros_like(detection_scores)
-        )).T,
-        columns=['onset', 'offset', 'a', 'b'],
-    )
-    change_point_scores, stats = accumulated_intermediate_statistics(
-        scores={'1': scores},
-        ground_truth={'1': []},
-        dtc_threshold=.5, gtc_threshold=.5, cttc_threshold=.5,
-    )[0]['a']
-    expected_change_point_scores = [4, np.inf]  # highest score where a false positive occurs when threshold falls below it
-    expected_true_positives = [0, 0]
-    expected_false_positives = [1, 0]
-    expected_cross_triggers = [0, 0]
-
-    assert stats['n_ref'] == 0, stats['n_ref']
-    assert np.abs(stats['t_ref']) < 1e-12, stats['t_ref']
-    assert len(change_point_scores) == len(expected_change_point_scores), (
-        change_point_scores, expected_change_point_scores)
-    assert (change_point_scores == expected_change_point_scores).all(), (
-        change_point_scores, expected_change_point_scores)
-    assert (
-        (stats['tps'] == expected_true_positives).all()
-    ), (stats['tps'], expected_true_positives)
-    assert (stats['fps'] == expected_false_positives).all(), (
-        stats['fps'], expected_false_positives)
-    assert stats['cts'].keys() == {'b'}, stats['cts']
-    assert stats['cts']['b'].shape == (len(change_point_scores),), (
-        stats['cts']['b'], expected_cross_triggers)
-    assert (
-        (stats['cts']['b'] == expected_cross_triggers).all()
-    ), (stats['cts']['b'], expected_cross_triggers)
-
-
 @pytest.mark.parametrize("dtc_threshold", [.5, .6])
 def test_event_offset_beyond_file_offset(dtc_threshold):
     detection_scores = np.concatenate((np.arange(5), 1+np.arange(3)[::-1]))

tests/test_segment_based/test_segment_based_area_under_roc.py

@@ -9,10 +9,8 @@ def test_segment_based_area_under_roc_vs_sklearn():
     y_true = np.array([0, 0, 1, 1])
     y_scores = np.array([0.1, 0.4, 0.35, 0.8])
     timestamps = np.arange(len(y_scores)+1) * segment_length
+    gt = {'1': [(timestamps[2], timestamps[4], 'a')]}
     auroc_sklearn = roc_auc_score(y_true, y_scores)
-    gt = {
-        '1': [(timestamps[idx], timestamps[idx+1], 'a') for idx, t in enumerate(y_true) if t]
-    }
     scores = {
         '1': create_score_dataframe(
             y_scores[..., None], timestamps=timestamps, event_classes=['a']

tests/test_segment_based/test_segment_based_average_precision.py

@@ -9,10 +9,8 @@ def test_segment_based_average_precision_vs_sklearn():
     y_true = np.array([0, 0, 1, 1])
     y_scores = np.array([0.1, 0.4, 0.35, 0.8])
     timestamps = np.arange(len(y_scores)+1) * segment_length
+    gt = {'1': [(timestamps[2], timestamps[4], 'a')]}
     ap_sklearn = average_precision_score(y_true, y_scores)
-    gt = {
-        '1': [(timestamps[idx], timestamps[idx+1], 'a') for idx, t in enumerate(y_true) if t]
-    }
     scores = {
         '1': create_score_dataframe(
             y_scores[..., None], timestamps=timestamps, event_classes=['a']