apunkt/ai-privacy-toolkit
1
0
Fork 0
mirror of https://github.com/IBM/ai-privacy-toolkit.git synced 2026-07-11 16:22:10 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

Address review comments, add additional distribution comparison tests and make them externally configurable too, in addition to the alpha becoming configurable.

Browse source 
Signed-off-by: Maya Anderson <mayaa@il.ibm.com>
This commit is contained in:
Maya Anderson 2023-09-20 19:44:54 +03:00
parent 0ee0bf05d6
commit 34de3ff93b
7 changed files with 234 additions and 165 deletions
Download patch file Download diff file Expand all files Collapse all files

16
tests/test_data_assessment.py
View file

@ -21,6 +21,7 @@ MIN_SHARE = 0.5
MIN_ROC_AUC = 0.0
MIN_PRECISION = 0.0
NUM_SYNTH_SAMPLES = 100
NUM_SYNTH_COMPONENTS = 4
iris_dataset_np = get_iris_dataset_np()
@ -28,7 +29,7 @@ diabetes_dataset_np = get_diabetes_dataset_np()
nursery_dataset_pd = get_nursery_dataset_pd()
adult_dataset_pd = get_adult_dataset_pd()
mgr = DatasetAssessmentManager(DatasetAssessmentManagerConfig(persist_reports=True, generate_plots=False))
mgr = DatasetAssessmentManager(DatasetAssessmentManagerConfig(persist_reports=False, generate_plots=False))
def teardown_function():
@ -36,10 +37,10 @@ def teardown_function():
mgr.dump_all_scores_to_files()
anon_testdata = [('iris_np', iris_dataset_np, 'np', k, mgr) for k in range(2, 10, 4)] \
+ [('diabetes_np', diabetes_dataset_np, 'np', k, mgr) for k in range(2, 10, 4)] \
+ [('nursery_pd', nursery_dataset_pd, 'pd', k, mgr) for k in range(2, 10, 4)] \
+ [('adult_pd', adult_dataset_pd, 'pd', k, mgr) for k in range(2, 10, 4)]
anon_testdata = ([('iris_np', iris_dataset_np, 'np', k, mgr) for k in range(2, 10, 4)]
+ [('diabetes_np', diabetes_dataset_np, 'np', k, mgr) for k in range(2, 10, 4)]
+ [('nursery_pd', nursery_dataset_pd, 'pd', k, mgr) for k in range(2, 10, 4)]
+ [('adult_pd', adult_dataset_pd, 'pd', k, mgr) for k in range(2, 10, 4)])
@pytest.mark.parametrize("name, data, dataset_type, k, mgr", anon_testdata)
@ -97,13 +98,12 @@ def test_risk_kde(name, data, dataset_type, mgr):
else:
raise ValueError('Pandas dataset missing a preprocessing step')
num_synth_samples = x_train.shape[0] # required by the chi test
synth_data = ArrayDataset(
kde(num_synth_samples, n_components=num_synth_components, original_data=encoded))
kde(NUM_SYNTH_SAMPLES, n_components=num_synth_components, original_data=encoded))
original_data_members = ArrayDataset(encoded, y_train)
original_data_non_members = ArrayDataset(encoded_test, y_test)
dataset_name = 'kde' + str(num_synth_samples) + name
dataset_name = 'kde' + str(NUM_SYNTH_SAMPLES) + name
assess_privacy_and_validate_result(mgr, original_data_members, original_data_non_members, synth_data, dataset_name,
categorical_features)