def test_count_size_mismatch_error(self):
data = np.array([
np.array([0.1, -0.1, 0.1, 0.1] * 10),
np.array([0.1, -0.1, 0.1, 0.1] * 10)
])
selectors = [[]]
self.set_seed()
with self.assertRaises(DiffPrivSizeMismatch):
DiffPrivLaplaceSanitizer.count(data, selectors, self.epsilon)
def test_decompose_data_slice_invalid_decomposition_error(self):
data = np.array([0.1, -0.1, 0.01, 0.0, -0.01, 0.1, 0.1] * 10)
def selector_positive(data):
return data != 0.0
def selector_negative(data):
return data != 0.0
selectors = [
selector_positive,
selector_negative,
]
with self.assertRaises(DiffPrivInvalidDecomposition):
DiffPrivLaplaceSanitizer.decompose_data_slice(data, selectors)
def test_decompose_data_slice_three_selectors(self):
data = np.array([0.1, -0.1, 0.01, -0.01, 0.1, 0.1] * 10)
expected_result = np.array([(data > -0.1) &
(data < 0.1), data >= 0.1, data <= -0.1], )
def selector_near_zero(data):
return (data > -0.1) & (data < 0.1)
def selector_positive(data):
return data >= 0.1
def selector_negative(data):
return data <= -0.1
selectors = [
selector_near_zero,
selector_positive,
selector_negative,
]
result = DiffPrivLaplaceSanitizer.decompose_data_slice(data, selectors)
self.assertEqual(len(result), len(selectors))
for index in range(len(expected_result)):
np.testing.assert_almost_equal(result[index],
expected_result[index])
def test_count_multiple_with_postprocess_with_axis_1(self):
data = np.array([
np.array([0.1, -0.1, 0.1, 0.1] * 10),
np.array([0.1, -0.1, 0.1, 0.1] * 10),
])
postprocess = True
expected_result = [
np.array([np.sum(data[0] >= 0),
np.sum(data[0] < 0)]),
np.array([np.sum(data[1] >= 0),
np.sum(data[1] < 0)]),
]
def selector_positive(data):
return data >= 0
def selector_negative(data):
return data < 0
selectors = [
[selector_positive, selector_negative],
[selector_positive, selector_negative],
]
self.set_seed()
results = DiffPrivLaplaceSanitizer.count(data,
selectors,
self.epsilon,
axis=1,
postprocess=postprocess)
for index in range(len(expected_result)):
np.testing.assert_almost_equal(results[index],
expected_result[index])
def test_count_single_no_postprocess_with_axis_1(self):
data = np.array([0.1, -0.1, 0.1, 0.1] * 10)
postprocess = False
expected_result = [np.array([np.sum(data >= 0), np.sum(data < 0)])]
def selector_positive(data):
return data >= 0
def selector_negative(data):
return data < 0
selectors = [
selector_positive,
selector_negative,
]
data = data.transpose()
self.set_seed()
results = DiffPrivLaplaceSanitizer.count(data,
selectors,
self.epsilon,
axis=1,
postprocess=postprocess)
for index in range(len(expected_result)):
np.testing.assert_almost_equal(results[index],
expected_result[index])
def test_count_invalid_dimension_error(self):
data = np.array([np.array([np.array([0.1, -0.1, 0.1, 0.1] * 10)])])
def selector_positive(data):
return data >= 0
def selector_negative(data):
return data < 0
selectors = [
selector_positive,
selector_negative,
]
self.set_seed()
with self.assertRaises(DiffPrivInvalidDimensions):
DiffPrivLaplaceSanitizer.count(data, selectors, self.epsilon)
def test_count_example(self):
epsilon = 0.1
data = np.array([0.01, -0.01, 0.03, -0.001, 0.1] * 2)
expected_value = np.array([np.array([5.0, 5.0])])
def selector_positive(data):
return data >= 0.0
def selector_negative(data):
return data < 0.0
selectors = [selector_positive, selector_negative]
self.set_seed()
value = DiffPrivLaplaceSanitizer.count(data, selectors, epsilon)
np.testing.assert_almost_equal(value, expected_value)