import inspect
import os
from csrank.util import configure_logging_numpy_keras
if __name__ == '__main__':
DIR_PATH = os.path.dirname(
os.path.abspath(inspect.getfile(inspect.currentframe())))
log_path = os.path.join(DIR_PATH, 'logs', 'test.log')
configure_logging_numpy_keras(log_path=log_path, name="Test")
if cindex <= 0:
folder = "{}_single_fold"
result_folder = "single_cv_results"
else:
folder = "{}_multiple_folds"
result_folder = "multiple_cv_results"
file_name_format.format(file_name_format, cindex)
log_path = os.path.join(
DIR_PATH, folder.format('logs'),
file_name_format.format(dataset_str, ranker_name) + '.log')
random_state = np.random.RandomState(seed=seed)
create_dir_recursively(log_path, True)
# log_path = rename_file_if_exist(log_path)
logger = configure_logging_numpy_keras(seed=random_state.randint(2**32),
log_path=log_path)
logger.debug(arguments)
dataset_function_params['random_state'] = random_state
ranker, dataset_reader = get_ranker_and_dataset_functions(
ranker_name, dataset_name, dataset_function_params, problem)
X_train, Y_train, X_test, Y_test = dataset_reader.get_single_train_test_split(
)
n_features, n_objects = log_test_train_data(X_train, X_test, logger)
ranker_params, fit_params, parameter_ranges = get_ranker_parameters(
ranker_name, n_features, n_objects, dataset_name,
dataset_function_params)
logger.info("ranker_params {} fit_params {}".format(
ranker_params, fit_params))
# Finally, fit a model on the complete training set:
if __name__ == '__main__':
arguments = docopt(__doc__)
n_objects = int(arguments['--n_objects'])
dirname = os.path.dirname(os.path.abspath(inspect.getfile(inspect.currentframe())))
log_path = os.path.join(dirname, "logs", "generalizing_mean_{}.log".format(n_objects))
df_path = os.path.join(dirname, "logs", "generalizing_mean_{}.csv".format(n_objects))
log_path = rename_file_if_exist(log_path)
df_path = rename_file_if_exist(df_path)
random_state = np.random.RandomState(seed=42)
seed = random_state.randint(2 ** 32)
rows_list = []
logger = configure_logging_numpy_keras(seed=seed, log_path=log_path)
X_train, Y_train, _, _ = generate_medoid_dataset(n_objects=n_objects, random_state=seed)
n_instances, n_objects, n_features = X_train.shape
epochs = 1000
params = {"n_objects": n_objects, "n_object_features": n_features, "use_early_stopping": True}
logger.info("############################# With Default Set layers ##############################")
gor = FATEObjectRanker(n_hidden_set_layers=2, n_hidden_set_units=2, **params)
result = get_evaluation_result(gor, X_train, Y_train, epochs)
result[MODEL] = "2SetLayersDefaultParams"
rows_list.append(result)
logger.info("############################# With 32 set layers ##############################")
gor = FATEObjectRanker(n_hidden_set_layers=32, n_hidden_set_units=32, **params)
result = get_evaluation_result(gor, X_train, Y_train, epochs)