Function test_randomized_search

catboost/python-package/ut/medium/test.py:3945–3979

(task_type)

def test_randomized_search(task_type):
    pool = Pool(TRAIN_FILE, column_description=CD_FILE)
    model = CatBoost(
        {
            "learning_rate": 0.03,
            "loss_function": "Logloss",
            "eval_metric": "AUC",
            "task_type": task_type,
            "gpu_ram_part": TEST_GPU_RAM_PART,
        }
    )
    feature_border_type_list = ['Median', 'Uniform', 'UniformAndQuantiles', 'MaxLogSum']
    one_hot_max_size_list = [4, 7, 10]
    iterations_list = [5, 7, 10]
    border_count_list = [4, 10, 50, 100]
    results = model.randomized_search(
        {
            'feature_border_type': feature_border_type_list,
            'one_hot_max_size': one_hot_max_size_list,
            'iterations': iterations_list,
            'border_count': border_count_list
        },
        pool
    )
    assert "train-Logloss-mean" in results['cv_results'], '"train-Logloss-mean" not in results'

    prev_value = results['cv_results']["train-Logloss-mean"][0]
    for value in results['cv_results']["train-Logloss-mean"][1:]:
        assert value < prev_value, 'not monotonic Logloss-mean'
        prev_value = value

    assert results['params'].get('feature_border_type') in feature_border_type_list
    assert results['params'].get('one_hot_max_size') in one_hot_max_size_list
    assert results['params'].get('iterations') in iterations_list
    assert results['params'].get('border_count') in border_count_list


def test_randomized_search_only_dist(task_type):