def __init__(self,
threshold,
eps,
delta,
sampling_method,
pickle_dir,
print_progress=True):
self.sample = private_sampling.PrivateThresholdSampleWithFrequencies(
threshold, eps, delta, sampling_method, store_every=1)
file_name = "private_sampling_%s_t%s_e%s_d%s" % (
sampling_method.__name__, threshold, eps, 1 / delta)
self.file_path = os.path.join(pickle_dir, file_name)
self.print_progress = print_progress
if os.path.exists(self.file_path):
self._load()
def test_valid_reported_frequency_distribution(self):
"""Checks that the distribution of reported frequencies is valid.
Computes the distribution of frequencies that are reported (when computing
a private sample) and checks that it is valid: all probabilities are between
0 and 1, and they sum up to 1.
"""
s = private_sampling.PrivateThresholdSampleWithFrequencies(
threshold=0.5, eps=0.1, delta=0.5**20)
freq_dists = [
s.compute_reported_frequency_dist(i) for i in range(100, 1001, 100)
]
for dist in freq_dists:
self.assertAlmostEqual(sum(dist.values()), 1.0)
for x in dist.values():
self.assertGreaterEqual(x, 0.0)