def test_apply_kind_on_data_slice_multiple_axis_0_skip(self):
data = np.array([list(range(0, 20)) + [100.0]] * 3)
kinds = [DiffPrivStatisticKind.all, None, DiffPrivStatisticKind.all]
expected_results = [
self.calculate_stats(data[0]),
None,
self.calculate_stats(data[2]),
]
data = np.transpose(data)
self.set_seed()
results = DiffPrivStatistics.apply_kind_on_data_slice(
data, kinds, self.epsilon, axis=0
)
self.assertEqual(len(results), len(expected_results))
for index in range(len(expected_results)):
result = results[index]
expected_result = expected_results[index]
if expected_result:
self.assertEqual(len(result), len(expected_result))
for key, value in expected_result.items():
value = result[key]
expected_value = expected_result[key]
self.assertAlmostEqual(value, expected_value, self.decimal_places)
else:
self.assertEqual(result, expected_result)
def test_proportion_multiple(self):
data = np.array([[True, False, True, True] * 10] * 3)
expected_values = np.count_nonzero(data, axis=1)
expected_values = np.divide(expected_values, np.size(data, axis=1))
self.set_seed()
value = DiffPrivStatistics.proportion(data, self.epsilon, axis=1)
np.testing.assert_almost_equal(value, expected_values, self.decimal_places)
def test_proportion_single(self):
data = np.array([True, False, True, True] * 10)
expected_value = np.count_nonzero(data)
expected_value = np.divide(expected_value, np.size(data))
self.set_seed()
value = DiffPrivStatistics.proportion(data, self.epsilon)
self.assertAlmostEqual(value, expected_value, self.decimal_places)
def test_median_multiple(self):
data = np.array([list(range(0, 20)) + [100.0]] * 3)
expected_values = np.median(data, axis=1)
self.set_seed()
value = DiffPrivStatistics.median(data, self.epsilon, axis=1)
np.testing.assert_almost_equal(value, expected_values, self.decimal_places)
def test_min_single(self):
data = list(range(3, 20))
data.reverse()
data = np.array(data)
expected_value = np.min(data)
self.set_seed()
value = DiffPrivStatistics.min(data, self.epsilon)
self.assertAlmostEqual(value, expected_value, self.decimal_places)
def test_min_multiple(self):
data = list(range(3, 20))
data.reverse()
data = np.array([data] * 3)
expected_values = np.min(data, axis=1)
self.set_seed()
value = DiffPrivStatistics.min(data, self.epsilon, axis=1)
np.testing.assert_almost_equal(value, expected_values, self.decimal_places)
def test_count_single_with_condition(self):
data = np.array([0.1, -0.1, 0.1, 0.1] * 10)
expected_value = np.count_nonzero(data >= 0)
def condition(data):
return data >= 0.0
self.set_seed()
value = DiffPrivStatistics.count(data, self.epsilon, condition=condition)
self.assertAlmostEqual(value, expected_value, self.decimal_places)
def test_apply_kind_on_data_slice_single_skip(self):
data = np.array(list(range(0, 20)) + [100.0])
kinds = None
expected_results = [None]
self.set_seed()
results = DiffPrivStatistics.apply_kind_on_data_slice(data, kinds, self.epsilon)
self.assertEqual(len(results), len(expected_results))
for index in range(len(expected_results)):
result = results[index]
expected_result = expected_results[index]
self.assertEqual(result, expected_result)
def test_count_multiple_with_condition(self):
data = np.array([[0.1, -0.1, 0.1, 0.1] * 10] * 3)
expected_values = np.count_nonzero(data >= 0, axis=1)
def condition(data):
return data >= 0.0
self.set_seed()
value = DiffPrivStatistics.count(
data, self.epsilon, condition=condition, axis=1
)
np.testing.assert_almost_equal(value, expected_values, self.decimal_places)
def test_apply_kind_on_data_slice_single(self):
data = np.array(list(range(0, 20)) + [100.0])
kinds = DiffPrivStatisticKind.all
expected_results = [self.calculate_stats(data)]
self.set_seed()
results = DiffPrivStatistics.apply_kind_on_data_slice(data, kinds, self.epsilon)
self.assertEqual(len(results), len(expected_results))
for index in range(len(expected_results)):
result = results[index]
expected_result = expected_results[index]
self.assertEqual(len(result), len(expected_result))
for key, value in expected_result.items():
value = result[key]
expected_value = expected_result[key]
self.assertAlmostEqual(value, expected_value, self.decimal_places)
def test_count_single(self):
data = np.array([True, False, True, True] * 10)
expected_value = np.count_nonzero(data)
self.set_seed()
value = DiffPrivStatistics.count(data, self.epsilon)
self.assertAlmostEqual(value, expected_value, self.decimal_places)
def test_apply_kind_on_data_slice_size_mismatch_error(self):
data = np.array(list(range(0, 20)) + [100.0])
kinds = [DiffPrivStatisticKind.all, DiffPrivStatisticKind.all]
self.set_seed()
with self.assertRaises(DiffPrivSizeMismatch):
DiffPrivStatistics.apply_kind_on_data_slice(data, kinds, self.epsilon)
def test_apply_kind_on_data_slice_invalid_dimension_error(self):
data = np.array([[list(range(0, 20)) + [100.0]]])
kinds = DiffPrivStatisticKind.all
self.set_seed()
with self.assertRaises(DiffPrivInvalidDimensions):
DiffPrivStatistics.apply_kind_on_data_slice(data, kinds, self.epsilon)
def test_variance_single(self):
data = np.array(list(range(0, 20)) + [100.0])
expected_value = np.var(data)
self.set_seed()
value = DiffPrivStatistics.variance(data, self.epsilon)
self.assertAlmostEqual(value, expected_value, self.decimal_places)