def _test_config_44(data_rnd_seed_values,
noise_values,
results_dir="results/synthetic_data"):
"""Testing configuration for the changing amount of noise experiment.
Parameters
----------
data_rnd_seed_values : list
Values of data_rnd_seed for which to repeat the experiment.
noise_values : list
Values of the amount of noise for which to repeat the experiment.
results_dir : str
Directory in which to put the chaning_noise directory with the relevant
experiment results.
"""
# parameters of the synthetic Boolean MTL problem
attributes = 12
disjunct_degree = 6
n = 50
task_groups = 5
tasks_per_group = 5
noise = 0.0
n_learning_sets = 10
# parameters of the MTL problem tester
rnd_seed = 51
# dynamic parameters of the synthetic Boolean MTL problem
for data_rnd_seed in data_rnd_seed_values:
results_path_fmt = os.path.join(
path_prefix, results_dir,
"changing_noise/bool_func-a{}d{}n{}g{}tg{}"
"nse{{}}rs{}nls{}-seed{}-complete_test".format(
attributes, disjunct_degree, n, task_groups, tasks_per_group,
data_rnd_seed, n_learning_sets, rnd_seed))
if "run" in mode:
for noise in noise_values:
# prepare directories and loggers
results_path = results_path_fmt.format(noise)
if not os.path.exists(results_path):
os.makedirs(results_path)
log_file = os.path.join(results_path, "run-{}.log".\
format(time.strftime("%Y%m%d_%H%M%S")))
configure_logger(logger,
console_level=logging.INFO,
file_name=log_file)
log_base_learner_info(logger, base_learners_bool)
# generate boolean data with complete test sets
funcs_pickle_path = os.path.join(results_path,
"boolean_funcs.pkl")
tasks_data, tasks_complete_test_sets = \
synthetic_data.generate_boolean_data_with_complete_test_sets(
attributes, disjunct_degree, n, task_groups,
tasks_per_group, noise,
random_seed=data_rnd_seed,
n_learning_sets=n_learning_sets,
funcs_pickle_path=funcs_pickle_path)
# test the generated MTL problem
test_tasks(tasks_data,
results_path,
base_learners_bool,
measures_clas,
learners,
"pre-prepared_test",
rnd_seed=rnd_seed,
test=test,
unpickle=unpickle,
visualize=visualize,
preprepared_test_sets=tasks_complete_test_sets,
separate_figs=True,
cfg_logger=False)
if "combine" in mode:
combine_experiment_results(
results_path_fmt,
noise_values,
(results_path_fmt.format(pprint_iter(noise_values)) +
"-{}-{{}}.pdf".format(error_measure)),
n_learning_sets,
error_measure=error_measure,
title="Avg. results for tasks",
xlabel="% of noise")
def _test_config_44(data_rnd_seed_values, noise_values,
results_dir="results/synthetic_data"):
"""Testing configuration for the changing amount of noise experiment.
Parameters
----------
data_rnd_seed_values : list
Values of data_rnd_seed for which to repeat the experiment.
noise_values : list
Values of the amount of noise for which to repeat the experiment.
results_dir : str
Directory in which to put the chaning_noise directory with the relevant
experiment results.
"""
# parameters of the synthetic Boolean MTL problem
attributes = 12
disjunct_degree = 6
n = 50
task_groups = 5
tasks_per_group = 5
noise = 0.0
n_learning_sets = 10
# parameters of the MTL problem tester
rnd_seed = 51
# dynamic parameters of the synthetic Boolean MTL problem
for data_rnd_seed in data_rnd_seed_values:
results_path_fmt = os.path.join(path_prefix, results_dir,
"changing_noise/bool_func-a{}d{}n{}g{}tg{}"
"nse{{}}rs{}nls{}-seed{}-complete_test".format(attributes,
disjunct_degree, n, task_groups, tasks_per_group,
data_rnd_seed, n_learning_sets, rnd_seed))
if "run" in mode:
for noise in noise_values:
# prepare directories and loggers
results_path = results_path_fmt.format(noise)
if not os.path.exists(results_path):
os.makedirs(results_path)
log_file = os.path.join(results_path, "run-{}.log".\
format(time.strftime("%Y%m%d_%H%M%S")))
configure_logger(logger, console_level=logging.INFO,
file_name=log_file)
log_base_learner_info(logger, base_learners_bool)
# generate boolean data with complete test sets
funcs_pickle_path = os.path.join(results_path,
"boolean_funcs.pkl")
tasks_data, tasks_complete_test_sets = \
synthetic_data.generate_boolean_data_with_complete_test_sets(
attributes, disjunct_degree, n, task_groups,
tasks_per_group, noise,
random_seed=data_rnd_seed,
n_learning_sets=n_learning_sets,
funcs_pickle_path=funcs_pickle_path)
# test the generated MTL problem
test_tasks(tasks_data, results_path, base_learners_bool,
measures_clas, learners, "pre-prepared_test",
rnd_seed=rnd_seed, test=test,
unpickle=unpickle, visualize=visualize,
preprepared_test_sets=tasks_complete_test_sets,
separate_figs=True, cfg_logger=False)
if "combine" in mode:
combine_experiment_results(results_path_fmt,
noise_values,
(results_path_fmt.format(pprint_iter(noise_values)) +
"-{}-{{}}.pdf".format(error_measure)),
n_learning_sets, error_measure=error_measure,
title="Avg. results for tasks", xlabel="% of noise")
# prepare directories and loggers
results_path = (os.path.join(
path_prefix, "results/synthetic_data/"
"bool_func-a{}d{}n{}g{}tg{}nse{}rs{}nls{}-seed{}-complete_test"
"".format(attributes, disjunct_degree, n, task_groups,
tasks_per_group, noise, data_rnd_seed,
n_learning_sets, rnd_seed)))
if not os.path.exists(results_path):
os.makedirs(results_path)
log_file = os.path.join(
results_path,
"run-{}.log".format(time.strftime("%Y%m%d_%H%M%S")))
configure_logger(logger,
console_level=logging.INFO,
file_name=log_file)
log_base_learner_info(logger, base_learners_bool)
# generate boolean data with complete test sets
funcs_pickle_path = os.path.join(results_path, "boolean_funcs.pkl")
tasks_data, tasks_complete_test_sets = \
synthetic_data.generate_boolean_data_with_complete_test_sets(
attributes, disjunct_degree, n, task_groups,
tasks_per_group, noise, random_seed=data_rnd_seed,
n_learning_sets=n_learning_sets,
funcs_pickle_path=funcs_pickle_path)
# test the generated MTL problem
test_tasks(tasks_data,
results_path,
base_learners_bool,
measures_clas,
learners,
"pre-prepared_test",
n_learning_sets = 3
# parameters of the MTL problem tester
rnd_seed = 51
# prepare directories and loggers
results_path = (os.path.join(path_prefix, "results/synthetic_data/"
"bool_func-a{}d{}n{}g{}tg{}nse{}rs{}nls{}-seed{}-complete_test"
"".format(attributes, disjunct_degree, n, task_groups,
tasks_per_group, noise, data_rnd_seed, n_learning_sets,
rnd_seed)))
if not os.path.exists(results_path):
os.makedirs(results_path)
log_file = os.path.join(results_path,
"run-{}.log".format(time.strftime("%Y%m%d_%H%M%S")))
configure_logger(logger, console_level=logging.INFO,
file_name=log_file)
log_base_learner_info(logger, base_learners_bool)
# generate boolean data with complete test sets
funcs_pickle_path = os.path.join(results_path, "boolean_funcs.pkl")
tasks_data, tasks_complete_test_sets = \
synthetic_data.generate_boolean_data_with_complete_test_sets(
attributes, disjunct_degree, n, task_groups,
tasks_per_group, noise, random_seed=data_rnd_seed,
n_learning_sets=n_learning_sets,
funcs_pickle_path=funcs_pickle_path)
# test the generated MTL problem
test_tasks(tasks_data, results_path, base_learners_bool,
measures_clas, learners, "pre-prepared_test",
rnd_seed=rnd_seed,
test=test, unpickle=unpickle, visualize=visualize,
preprepared_test_sets=tasks_complete_test_sets,
separate_figs=True, cfg_logger=False)
