def test_runtimes_ncobras():
clusterer = COBRAS(noise_probability=0.05, minimum_approximation_order=3, maximum_approximation_order=10,
certainty_threshold=0.95)
querier = ProbabilisticNoisyQuerier(None, None, 0.05, 100)
tasks = make_n_run_10_fold_cross_validation("ncobras_0.05noise_runtimes", clusterer, querier,
Dataset.get_standard_dataset_names(), 3, cobras_result_extractor)
execute_list_of_clustering_tasks(tasks, tests_per_batch=200)
def NPU_MPCKMeans_VaryingAmountsOfNoise():
print("making tests")
tests = TestCollection()
query_budget = 200
for noise_percentage in [-1, 0.05, 0.10, 0.20]:
noise_text = str(noise_percentage) if noise_percentage != -1 else "no"
tests.add_10_times_10_fold_test(
"NPU_MPCKmeans_{}_noise_budget{}".format(noise_text, query_budget),
"NPU_MPCKmeans", mpck_means_algorithm_settings_to_string(),
Dataset.get_standard_dataset_names(), "probability_noise_querier",
probabilistic_noisy_querier_settings_to_string(
noise_percentage, query_budget))
run_tests_over_SSH_on_machines(tests, MACHINES_TO_USE)
comparison_name = "NPU_MPCKmeans_varying_amounts_of_noise"
test_names = [
"NPU_MPCKmeans_no_noise_budget200",
"NPU_MPCKmeans_0.05_noise_budget200",
"NPU_MPCKmeans_0.10_noise_budget200",
"NPU_MPCKmeans_0.20_noise_budget200"
]
line_names = [
"NPU_MPCKmeans_no_noise", "NPU_MPCKmeans_0.05_noise",
"NPU_MPCKmeans_0.10_noise", "NPU_MPCKmeans_0.20_noise"
]
calculate_aris_and_compare_for_tests(comparison_name,
test_names,
line_names,
query_budget=query_budget)
def test_varying_amounts_of_noise_cobras():
test_collection = []
# plain COBRAS
clusterer = COBRAS(correct_noise=False)
dataset_names = Dataset.get_standard_dataset_names()
nb_of_runs = 3
test_collection.extend(
test_varying_amounts_of_noise("new_cobras", clusterer, cobras_result_extractor, dataset_names=dataset_names,
query_budget=200, nb_of_runs=nb_of_runs))
execute_list_of_clustering_tasks(test_collection, tests_per_batch=200)
def test_varying_amounts_of_noise_NPU_COSC():
from generate_clusterings.algorithms.my_cosc import MyCOSCMatlab
from generate_clusterings.algorithms.my_npu import NPU
test_collection = []
clusterer = NPU(MyCOSCMatlab(run_fast_version=True), debug=True)
# dataset_names = ['hepatitis']
dataset_names = Dataset.get_standard_dataset_names()
dataset_names.remove("hepatitis")
nb_of_runs = 3
test_collection.extend(
test_varying_amounts_of_noise("new_NPU_COSC_fast", clusterer, cosc_result_extractor, dataset_names=dataset_names,
query_budget=200, nb_of_runs=nb_of_runs))
execute_list_of_clustering_tasks(test_collection, tests_per_batch=200)
def cobras_minimal_vs_all_cycles_test():
print("making tests")
tests = TestCollection()
tests.add_10_times_10_fold_test(
"ncobras_minimal_cycles", "COBRAS",
cobras_algorithm_settings_to_string(0.10, 3, 7, 0.91, 0.91, True,
False),
Dataset.get_standard_dataset_names(), "probability_noise_querier",
probabilistic_noisy_querier_settings_to_string(0.10, 25))
tests.add_10_times_10_fold_test(
"ncobras_all_cycles", "COBRAS",
cobras_algorithm_settings_to_string(0.10, 3, 7, 0.91, 0.91, True,
True),
Dataset.get_standard_dataset_names(), "probability_noise_querier",
probabilistic_noisy_querier_settings_to_string(0.10, 25))
run_tests_over_SSH_on_machines(tests, nb_of_computers=5)
comparison_name = "ncobras_minimal_vs_all_cycles"
test_names = ["ncobras_minimal_cycles", "ncobras_all_cycles"]
line_names = None
calculate_aris_and_compare_for_tests(comparison_name, test_names,
line_names)
def test_varying_amounts_of_noise(test_name_raw, clusterers, result_extractor, query_budget=200, nb_of_runs=3,
noise_percentages=(-1, 0.05, 0.1),
dataset_names=Dataset.get_standard_dataset_names()):
if not isinstance(clusterers, list):
clusterers = [clusterers for _ in range(3)]
assert len(clusterers) == len(noise_percentages)
test_collection = []
for clusterer, noise_percentage in zip(clusterers, noise_percentages):
noise_text = str(noise_percentage) if noise_percentage != -1 else "no"
test_name = f"{test_name_raw}_{noise_text}_budget_{query_budget}"
querier = ProbabilisticNoisyQuerier(None, None, noise_percentage, query_budget)
additional_tasks = make_n_run_10_fold_cross_validation(test_name, clusterer, querier, dataset_names, nb_of_runs,
result_extractor)
test_collection.extend(additional_tasks)
return test_collection
def test_varying_amounts_of_noise_ncobras():
test_collection = []
dataset_names = Dataset.get_standard_dataset_names()
nb_of_runs = 3
# nCOBRAS
clusterers = [COBRAS(noise_probability=0.05, minimum_approximation_order=3, maximum_approximation_order=8,
certainty_threshold=0.95),
COBRAS(noise_probability=0.05, minimum_approximation_order=3, maximum_approximation_order=8,
certainty_threshold=0.95),
COBRAS(noise_probability=0.10, minimum_approximation_order=3, maximum_approximation_order=8,
certainty_threshold=0.95)]
test_collection.extend(
test_varying_amounts_of_noise("new_ncobras", clusterers, cobras_result_extractor, dataset_names=dataset_names,
query_budget=200, nb_of_runs=nb_of_runs))
execute_list_of_clustering_tasks(test_collection, tests_per_batch=200)
