def test_file_output(self):
output_dir = os.path.join(os.getcwd(), '.test')
try:
shutil.rmtree(output_dir)
except Exception:
pass
X_train, Y_train, X_test, Y_test = get_dataset('boston')
X_valid = X_test[:25, ]
Y_valid = Y_test[:25, ]
X_test = X_test[25:, ]
Y_test = Y_test[25:, ]
D = Dummy()
D.info = {
'metric': R2_METRIC,
'task': REGRESSION,
'is_sparse': False,
'label_num': 3
}
D.data = {
'X_train': X_train,
'Y_train': Y_train,
'X_valid': X_valid,
'X_test': X_test
}
D.feat_type = ['numerical', 'Numerical', 'numerical', 'numerical']
D.name = 'test'
configuration_space = get_configuration_space(D.info)
while True:
configuration = configuration_space.sample_configuration()
evaluator = HoldoutEvaluator(D,
configuration,
with_predictions=True,
all_scoring_functions=True,
output_dir=output_dir,
output_y_test=True)
if not self._fit(evaluator):
continue
evaluator.predict()
evaluator.file_output()
self.assertTrue(
os.path.exists(
os.path.join(output_dir, '.auto-sklearn',
'true_targets_ensemble.npy')))
break
def test_file_output(self):
output_dir = os.path.join(os.getcwd(), '.test')
try:
shutil.rmtree(output_dir)
except Exception:
pass
X_train, Y_train, X_test, Y_test = get_dataset('boston')
X_valid = X_test[:25, ]
Y_valid = Y_test[:25, ]
X_test = X_test[25:, ]
Y_test = Y_test[25:, ]
D = Dummy()
D.info = {
'metric': R2_METRIC,
'task': REGRESSION,
'is_sparse': False,
'label_num': 3
}
D.data = {
'X_train': X_train,
'Y_train': Y_train,
'X_valid': X_valid,
'X_test': X_test
}
D.feat_type = ['numerical', 'Numerical', 'numerical', 'numerical']
D.name = 'test'
configuration_space = get_configuration_space(D.info)
while True:
configuration = configuration_space.sample_configuration()
evaluator = HoldoutEvaluator(D, configuration,
with_predictions=True,
all_scoring_functions=True,
output_dir=output_dir,
output_y_test=True)
if not self._fit(evaluator):
continue
evaluator.predict()
evaluator.file_output()
self.assertTrue(os.path.exists(os.path.join(
output_dir, '.auto-sklearn', 'true_targets_ensemble.npy')))
break
def test_evaluate_multiclass_classification_all_metrics(self):
X_train, Y_train, X_test, Y_test = get_dataset("iris")
X_valid = X_test[:25,]
Y_valid = Y_test[:25,]
X_test = X_test[25:,]
Y_test = Y_test[25:,]
D = Dummy()
D.info = {"metric": BAC_METRIC, "task": MULTICLASS_CLASSIFICATION, "is_sparse": False, "label_num": 3}
D.data = {"X_train": X_train, "Y_train": Y_train, "X_valid": X_valid, "X_test": X_test}
D.feat_type = ["numerical", "Numerical", "numerical", "numerical"]
configuration_space = get_configuration_space(D.info, include_estimators=["lda"], include_preprocessors=["pca"])
# Test all scoring functions
err = []
for i in range(N_TEST_RUNS):
print("Evaluate configuration: %d; result:" % i)
configuration = configuration_space.sample_configuration()
D_ = copy.deepcopy(D)
evaluator = HoldoutEvaluator(D_, configuration, all_scoring_functions=True)
if not self._fit(evaluator):
continue
err.append(evaluator.predict())
print(err[-1])
self.assertIsInstance(err[-1], dict)
for key in err[-1]:
self.assertEqual(len(err[-1]), 5)
self.assertTrue(np.isfinite(err[-1][key]))
self.assertGreaterEqual(err[-1][key], 0.0)
print("Number of times it was worse than random guessing:" + str(np.sum(err > 1)))
def test_file_output(self):
output_dir = os.path.join(os.getcwd(), ".test")
try:
shutil.rmtree(output_dir)
except Exception:
pass
X_train, Y_train, X_test, Y_test = get_dataset("boston")
X_valid = X_test[:25,]
Y_valid = Y_test[:25,]
X_test = X_test[25:,]
Y_test = Y_test[25:,]
D = Dummy()
D.info = {"metric": R2_METRIC, "task": REGRESSION, "is_sparse": False, "label_num": 3}
D.data = {"X_train": X_train, "Y_train": Y_train, "X_valid": X_valid, "X_test": X_test}
D.feat_type = ["numerical", "Numerical", "numerical", "numerical"]
D.name = "test"
configuration_space = get_configuration_space(D.info)
while True:
configuration = configuration_space.sample_configuration()
evaluator = HoldoutEvaluator(
D,
configuration,
with_predictions=True,
all_scoring_functions=True,
output_dir=output_dir,
output_y_test=True,
)
if not self._fit(evaluator):
continue
evaluator.predict()
evaluator.file_output()
self.assertTrue(os.path.exists(os.path.join(output_dir, ".auto-sklearn", "true_targets_ensemble.npy")))
break
def test_evaluate_binary_classification(self):
X_train, Y_train, X_test, Y_test = get_dataset('iris')
eliminate_class_two = Y_train != 2
X_train = X_train[eliminate_class_two]
Y_train = Y_train[eliminate_class_two]
eliminate_class_two = Y_test != 2
X_test = X_test[eliminate_class_two]
Y_test = Y_test[eliminate_class_two]
X_valid = X_test[:25, ]
Y_valid = Y_test[:25, ]
X_test = X_test[25:, ]
Y_test = Y_test[25:, ]
D = Dummy()
D.info = {
'metric': AUC_METRIC,
'task': BINARY_CLASSIFICATION,
'is_sparse': False,
'label_num': 2
}
D.data = {
'X_train': X_train,
'Y_train': Y_train,
'X_valid': X_valid,
'X_test': X_test
}
D.feat_type = ['numerical', 'Numerical', 'numerical', 'numerical']
configuration_space = get_configuration_space(
D.info,
include_estimators=['lda'],
include_preprocessors=['pca'])
err = np.zeros([N_TEST_RUNS])
for i in range(N_TEST_RUNS):
print('Evaluate configuration: %d; result:' % i)
configuration = configuration_space.sample_configuration()
D_ = copy.deepcopy(D)
evaluator = HoldoutEvaluator(D_, configuration)
if not self._fit(evaluator):
continue
err[i] = evaluator.predict()
self.assertTrue(np.isfinite(err[i]))
print(err[i])
self.assertGreaterEqual(err[i], 0.0)
def test_evaluate_binary_classification(self):
X_train, Y_train, X_test, Y_test = get_dataset('iris')
eliminate_class_two = Y_train != 2
X_train = X_train[eliminate_class_two]
Y_train = Y_train[eliminate_class_two]
eliminate_class_two = Y_test != 2
X_test = X_test[eliminate_class_two]
Y_test = Y_test[eliminate_class_two]
X_valid = X_test[:25, ]
Y_valid = Y_test[:25, ]
X_test = X_test[25:, ]
Y_test = Y_test[25:, ]
D = Dummy()
D.info = {
'metric': AUC_METRIC,
'task': BINARY_CLASSIFICATION,
'is_sparse': False,
'label_num': 2
}
D.data = {
'X_train': X_train,
'Y_train': Y_train,
'X_valid': X_valid,
'X_test': X_test
}
D.feat_type = ['numerical', 'Numerical', 'numerical', 'numerical']
configuration_space = get_configuration_space(
D.info, include_estimators=['lda'], include_preprocessors=['pca'])
err = np.zeros([N_TEST_RUNS])
for i in range(N_TEST_RUNS):
print('Evaluate configuration: %d; result:' % i)
configuration = configuration_space.sample_configuration()
D_ = copy.deepcopy(D)
evaluator = HoldoutEvaluator(D_, configuration)
if not self._fit(evaluator):
continue
err[i] = evaluator.predict()
self.assertTrue(np.isfinite(err[i]))
print(err[i])
self.assertGreaterEqual(err[i], 0.0)
def test_evaluate_multiclass_classification_all_metrics(self):
X_train, Y_train, X_test, Y_test = get_dataset('iris')
X_valid = X_test[:25, ]
Y_valid = Y_test[:25, ]
X_test = X_test[25:, ]
Y_test = Y_test[25:, ]
D = Dummy()
D.info = {
'metric': BAC_METRIC,
'task': MULTICLASS_CLASSIFICATION,
'is_sparse': False,
'label_num': 3
}
D.data = {
'X_train': X_train,
'Y_train': Y_train,
'X_valid': X_valid,
'X_test': X_test
}
D.feat_type = ['numerical', 'Numerical', 'numerical', 'numerical']
configuration_space = get_configuration_space(
D.info,
include_estimators=['lda'],
include_preprocessors=['pca'])
# Test all scoring functions
err = []
for i in range(N_TEST_RUNS):
print('Evaluate configuration: %d; result:' % i)
configuration = configuration_space.sample_configuration()
D_ = copy.deepcopy(D)
evaluator = HoldoutEvaluator(D_, configuration,
all_scoring_functions=True)
if not self._fit(evaluator):
continue
err.append(evaluator.predict())
print(err[-1])
self.assertIsInstance(err[-1], dict)
for key in err[-1]:
self.assertEqual(len(err[-1]), 5)
self.assertTrue(np.isfinite(err[-1][key]))
self.assertGreaterEqual(err[-1][key], 0.0)
def test_evaluate_multiclass_classification_all_metrics(self):
X_train, Y_train, X_test, Y_test = get_dataset('iris')
X_valid = X_test[:25, ]
Y_valid = Y_test[:25, ]
X_test = X_test[25:, ]
Y_test = Y_test[25:, ]
D = Dummy()
D.info = {
'metric': BAC_METRIC,
'task': MULTICLASS_CLASSIFICATION,
'is_sparse': False,
'label_num': 3
}
D.data = {
'X_train': X_train,
'Y_train': Y_train,
'X_valid': X_valid,
'X_test': X_test
}
D.feat_type = ['numerical', 'Numerical', 'numerical', 'numerical']
configuration_space = get_configuration_space(
D.info, include_estimators=['lda'], include_preprocessors=['pca'])
# Test all scoring functions
err = []
for i in range(N_TEST_RUNS):
print('Evaluate configuration: %d; result:' % i)
configuration = configuration_space.sample_configuration()
D_ = copy.deepcopy(D)
evaluator = HoldoutEvaluator(D_,
configuration,
all_scoring_functions=True)
if not self._fit(evaluator):
continue
err.append(evaluator.predict())
print(err[-1])
self.assertIsInstance(err[-1], dict)
for key in err[-1]:
self.assertEqual(len(err[-1]), 5)
self.assertTrue(np.isfinite(err[-1][key]))
self.assertGreaterEqual(err[-1][key], 0.0)
def test_evaluate_regression(self):
X_train, Y_train, X_test, Y_test = get_dataset('boston')
X_valid = X_test[:200, ]
Y_valid = Y_test[:200, ]
X_test = X_test[200:, ]
Y_test = Y_test[200:, ]
D = Dummy()
D.info = {
'metric': R2_METRIC,
'task': REGRESSION,
'is_sparse': False,
'label_num': 1
}
D.data = {
'X_train': X_train,
'Y_train': Y_train,
'X_valid': X_valid,
'X_test': X_test
}
D.feat_type = [
'numerical', 'Numerical', 'numerical', 'numerical', 'numerical',
'numerical', 'numerical', 'numerical', 'numerical', 'numerical',
'numerical'
]
configuration_space = get_configuration_space(
D.info,
include_estimators=['extra_trees'],
include_preprocessors=['no_preprocessing'])
err = np.zeros([N_TEST_RUNS])
for i in range(N_TEST_RUNS):
print('Evaluate configuration: %d; result:' % i)
configuration = configuration_space.sample_configuration()
D_ = copy.deepcopy(D)
evaluator = HoldoutEvaluator(D_, configuration)
if not self._fit(evaluator):
continue
err[i] = evaluator.predict()
self.assertTrue(np.isfinite(err[i]))
print(err[i])
self.assertGreaterEqual(err[i], 0.0)
def test_evaluate_multilabel_classification(self):
X_train, Y_train, X_test, Y_test = get_dataset('iris')
Y_train = np.array(convert_to_bin(Y_train, 3))
Y_train[:, -1] = 1
Y_test = np.array(convert_to_bin(Y_test, 3))
Y_test[:, -1] = 1
X_valid = X_test[:25, ]
Y_valid = Y_test[:25, ]
X_test = X_test[25:, ]
Y_test = Y_test[25:, ]
D = Dummy()
D.info = {
'metric': F1_METRIC,
'task': MULTILABEL_CLASSIFICATION,
'is_sparse': False,
'label_num': 3
}
D.data = {
'X_train': X_train,
'Y_train': Y_train,
'X_valid': X_valid,
'X_test': X_test
}
D.feat_type = ['numerical', 'Numerical', 'numerical', 'numerical']
configuration_space = get_configuration_space(
D.info,
include_estimators=['extra_trees'],
include_preprocessors=['no_preprocessing'])
err = np.zeros([N_TEST_RUNS])
for i in range(N_TEST_RUNS):
print('Evaluate configuration: %d; result:' % i)
configuration = configuration_space.sample_configuration()
D_ = copy.deepcopy(D)
evaluator = HoldoutEvaluator(D_, configuration)
if not self._fit(evaluator):
continue
err[i] = evaluator.predict()
print(err[i])
self.assertTrue(np.isfinite(err[i]))
self.assertGreaterEqual(err[i], 0.0)
def test_evaluate_multilabel_classification(self):
X_train, Y_train, X_test, Y_test = get_dataset('iris')
Y_train = np.array(convert_to_bin(Y_train, 3))
Y_train[:, -1] = 1
Y_test = np.array(convert_to_bin(Y_test, 3))
Y_test[:, -1] = 1
X_valid = X_test[:25, ]
Y_valid = Y_test[:25, ]
X_test = X_test[25:, ]
Y_test = Y_test[25:, ]
D = Dummy()
D.info = {
'metric': F1_METRIC,
'task': MULTILABEL_CLASSIFICATION,
'is_sparse': False,
'label_num': 3
}
D.data = {
'X_train': X_train,
'Y_train': Y_train,
'X_valid': X_valid,
'X_test': X_test
}
D.feat_type = ['numerical', 'Numerical', 'numerical', 'numerical']
configuration_space = get_configuration_space(
D.info,
include_estimators=['extra_trees'],
include_preprocessors=['no_preprocessing'])
err = np.zeros([N_TEST_RUNS])
for i in range(N_TEST_RUNS):
print('Evaluate configuration: %d; result:' % i)
configuration = configuration_space.sample_configuration()
D_ = copy.deepcopy(D)
evaluator = HoldoutEvaluator(D_, configuration)
if not self._fit(evaluator):
continue
err[i] = evaluator.predict()
print(err[i])
self.assertTrue(np.isfinite(err[i]))
self.assertGreaterEqual(err[i], 0.0)
def test_evaluate_regression(self):
X_train, Y_train, X_test, Y_test = get_dataset("boston")
X_valid = X_test[:200,]
Y_valid = Y_test[:200,]
X_test = X_test[200:,]
Y_test = Y_test[200:,]
D = Dummy()
D.info = {"metric": R2_METRIC, "task": REGRESSION, "is_sparse": False, "label_num": 1}
D.data = {"X_train": X_train, "Y_train": Y_train, "X_valid": X_valid, "X_test": X_test}
D.feat_type = [
"numerical",
"Numerical",
"numerical",
"numerical",
"numerical",
"numerical",
"numerical",
"numerical",
"numerical",
"numerical",
"numerical",
]
configuration_space = get_configuration_space(
D.info, include_estimators=["extra_trees"], include_preprocessors=["no_preprocessing"]
)
err = np.zeros([N_TEST_RUNS])
for i in range(N_TEST_RUNS):
print("Evaluate configuration: %d; result:" % i)
configuration = configuration_space.sample_configuration()
D_ = copy.deepcopy(D)
evaluator = HoldoutEvaluator(D_, configuration)
if not self._fit(evaluator):
continue
err[i] = evaluator.predict()
self.assertTrue(np.isfinite(err[i]))
print(err[i])
self.assertGreaterEqual(err[i], 0.0)
print("Number of times it was worse than random guessing:" + str(np.sum(err > 1)))
def test_evaluate_regression(self):
X_train, Y_train, X_test, Y_test = get_dataset('boston')
X_valid = X_test[:200, ]
Y_valid = Y_test[:200, ]
X_test = X_test[200:, ]
Y_test = Y_test[200:, ]
D = Dummy()
D.info = {
'metric': R2_METRIC,
'task': REGRESSION,
'is_sparse': False,
'label_num': 1
}
D.data = {
'X_train': X_train,
'Y_train': Y_train,
'X_valid': X_valid,
'X_test': X_test
}
D.feat_type = ['numerical', 'Numerical', 'numerical', 'numerical',
'numerical', 'numerical', 'numerical', 'numerical',
'numerical', 'numerical', 'numerical']
configuration_space = get_configuration_space(
D.info,
include_estimators=['extra_trees'],
include_preprocessors=['no_preprocessing'])
err = np.zeros([N_TEST_RUNS])
for i in range(N_TEST_RUNS):
print('Evaluate configuration: %d; result:' % i)
configuration = configuration_space.sample_configuration()
D_ = copy.deepcopy(D)
evaluator = HoldoutEvaluator(D_, configuration)
if not self._fit(evaluator):
continue
err[i] = evaluator.predict()
self.assertTrue(np.isfinite(err[i]))
print(err[i])
self.assertGreaterEqual(err[i], 0.0)
def test_with_abalone(self):
dataset = "abalone"
dataset_path = os.path.join(os.path.dirname(__file__), ".datasets", dataset)
D = CompetitionDataManager(dataset_path)
configuration_space = get_configuration_space(
D.info, include_estimators=["extra_trees"], include_preprocessors=["no_preprocessing"]
)
errors = []
for i in range(N_TEST_RUNS):
configuration = configuration_space.sample_configuration()
D_ = copy.deepcopy(D)
evaluator = HoldoutEvaluator(D_, configuration)
if not self._fit(evaluator):
continue
err = evaluator.predict()
self.assertLess(err, 0.99)
self.assertTrue(np.isfinite(err))
errors.append(err)
# This is a reasonable bound
self.assertEqual(10, len(errors))
self.assertLess(min(errors), 0.77)
def test_evaluate_binary_classification(self):
X_train, Y_train, X_test, Y_test = get_dataset("iris")
eliminate_class_two = Y_train != 2
X_train = X_train[eliminate_class_two]
Y_train = Y_train[eliminate_class_two]
eliminate_class_two = Y_test != 2
X_test = X_test[eliminate_class_two]
Y_test = Y_test[eliminate_class_two]
X_valid = X_test[:25,]
Y_valid = Y_test[:25,]
X_test = X_test[25:,]
Y_test = Y_test[25:,]
D = Dummy()
D.info = {"metric": AUC_METRIC, "task": BINARY_CLASSIFICATION, "is_sparse": False, "label_num": 2}
D.data = {"X_train": X_train, "Y_train": Y_train, "X_valid": X_valid, "X_test": X_test}
D.feat_type = ["numerical", "Numerical", "numerical", "numerical"]
configuration_space = get_configuration_space(D.info, include_estimators=["lda"], include_preprocessors=["pca"])
err = np.zeros([N_TEST_RUNS])
for i in range(N_TEST_RUNS):
print("Evaluate configuration: %d; result:" % i)
configuration = configuration_space.sample_configuration()
D_ = copy.deepcopy(D)
evaluator = HoldoutEvaluator(D_, configuration)
if not self._fit(evaluator):
continue
err[i] = evaluator.predict()
self.assertTrue(np.isfinite(err[i]))
print(err[i])
self.assertGreaterEqual(err[i], 0.0)
print("Number of times it was worse than random guessing:" + str(np.sum(err > 1)))
def test_with_abalone(self):
dataset = 'abalone'
dataset_path = os.path.join(os.path.dirname(__file__), '.datasets',
dataset)
D = CompetitionDataManager(dataset_path)
configuration_space = get_configuration_space(
D.info,
include_estimators=['extra_trees'],
include_preprocessors=['no_preprocessing'])
errors = []
for i in range(N_TEST_RUNS):
configuration = configuration_space.sample_configuration()
D_ = copy.deepcopy(D)
evaluator = HoldoutEvaluator(D_, configuration)
if not self._fit(evaluator):
continue
err = evaluator.predict()
self.assertLess(err, 0.99)
self.assertTrue(np.isfinite(err))
errors.append(err)
# This is a reasonable bound
self.assertEqual(10, len(errors))
self.assertLess(min(errors), 0.77)
def test_evaluate_multilabel_classification(self):
X_train, Y_train, X_test, Y_test = get_dataset("iris")
Y_train = np.array(convert_to_bin(Y_train, 3))
Y_train[:, -1] = 1
Y_test = np.array(convert_to_bin(Y_test, 3))
Y_test[:, -1] = 1
X_valid = X_test[:25,]
Y_valid = Y_test[:25,]
X_test = X_test[25:,]
Y_test = Y_test[25:,]
D = Dummy()
D.info = {"metric": F1_METRIC, "task": MULTILABEL_CLASSIFICATION, "is_sparse": False, "label_num": 3}
D.data = {"X_train": X_train, "Y_train": Y_train, "X_valid": X_valid, "X_test": X_test}
D.feat_type = ["numerical", "Numerical", "numerical", "numerical"]
configuration_space = get_configuration_space(
D.info, include_estimators=["extra_trees"], include_preprocessors=["no_preprocessing"]
)
err = np.zeros([N_TEST_RUNS])
for i in range(N_TEST_RUNS):
print("Evaluate configuration: %d; result:" % i)
configuration = configuration_space.sample_configuration()
D_ = copy.deepcopy(D)
evaluator = HoldoutEvaluator(D_, configuration)
if not self._fit(evaluator):
continue
err[i] = evaluator.predict()
print(err[i])
self.assertTrue(np.isfinite(err[i]))
self.assertGreaterEqual(err[i], 0.0)
print("Number of times it was worse than random guessing:" + str(np.sum(err > 1)))
