Default number of cv_folds is changed to 5 (#865)

* changed default cv_folds number in params and some tests
aimclub · Sep 8, 2022 · c8eed58 · c8eed58
1 parent 6487966
commit c8eed58
Show file tree

Hide file tree

Showing 4 changed files with 6 additions and 6 deletions.
diff --git a/test/unit/composer/test_history.py b/test/unit/composer/test_history.py
@@ -164,7 +164,7 @@ def test_collect_intermediate_metric(pipeline: Pipeline, input_data: InputData,
 
 
 @pytest.mark.parametrize("cv_generator, data",
-                         [(partial(tabular_cv_generator, folds=3),
+                         [(partial(tabular_cv_generator, folds=5),
                            get_classification_data()),
                           (partial(ts_cv_generator, folds=3, validation_blocks=2),
                            get_ts_data()[0])])

diff --git a/test/unit/optimizer/test_data_objective_eval.py b/test/unit/optimizer/test_data_objective_eval.py
@@ -95,7 +95,7 @@ def test_pipeline_objective_evaluate_with_empty_pipeline(classification_dataset)
 def test_pipeline_objective_evaluate_with_cv_fold(classification_dataset):
     pipeline = sample_pipeline()
 
-    cv_fold = partial(tabular_cv_generator, classification_dataset, folds=3)
+    cv_fold = partial(tabular_cv_generator, classification_dataset, folds=5)
     metric = ClassificationMetricsEnum.logloss
 
     objective_eval = PipelineObjectiveEvaluate(Objective(metric), cv_fold)

diff --git a/test/unit/validation/test_table_cv.py b/test/unit/validation/test_table_cv.py
@@ -44,7 +44,7 @@ def get_classification_data():
 def test_cv_multiple_metrics_evaluated_correct(classification_dataset):
     pipeline = sample_pipeline()
 
-    cv_folds = partial(tabular_cv_generator, classification_dataset, folds=3)
+    cv_folds = partial(tabular_cv_generator, classification_dataset, folds=5)
     metrics = [ClassificationMetricsEnum.ROCAUC_penalty,
                ClassificationMetricsEnum.accuracy,
                ClassificationMetricsEnum.logloss]
@@ -80,7 +80,7 @@ def test_cv_min_kfolds_raise():
 
 
 def test_tuner_cv_classification_correct():
-    folds = 2
+    folds = 5
     dataset = get_iris_data()
 
     simple_pipeline = pipeline_simple()
@@ -105,7 +105,7 @@ def test_composer_with_cv_optimization_correct():
     composer_requirements = PipelineComposerRequirements(primary=available_model_types,
                                                          secondary=available_model_types,
                                                          timeout=timedelta(minutes=0.2),
-                                                         num_of_generations=2, cv_folds=3,
+                                                         num_of_generations=2, cv_folds=5,
                                                          show_progress=False)
 
     builder = ComposerBuilder(task).with_requirements(composer_requirements).with_metrics(metric_function)

diff --git a/test/unit/validation/test_time_series_cv.py b/test/unit/validation/test_time_series_cv.py
@@ -37,7 +37,7 @@ def test_ts_cv_generator_correct():
     """ Checks if the split into training and test for time series cross
     validation is correct
 
-    By default, the number of validation blocks for each fold is three
+    By default, the number of validation blocks for each fold is five
     """
     folds = 2
     forecast_len, validation_blocks, time_series = configure_experiment()