def Run(self, W, relation, eps, seed):
prng = np.random.RandomState(seed)
M = pselection.PrivBayesSelect(self.theta, self.domain,
eps).select(relation, prng)
x = transformation.Vectorize(
'', reduced_domain=self.domain).transform(relation)
y = measurement.Laplace(M, eps).measure(x, prng)
x_hat = inference.LeastSquares().infer(M, y)
return x_hat
import yaml
CSV_PATH = os.environ['EKTELO_DATA']
CONFIG_PATH = os.path.join(os.environ['EKTELO_HOME'], 'resources', 'config')
# Load relation
filename = os.path.join(CSV_PATH, 'cps.csv')
config_file = os.path.join(CONFIG_PATH, 'cps.yml')
config = yaml.load(open(config_file, 'r').read())['cps_config']
R = data.Relation(config).load_csv(filename, ',')
# Choose reduced domain for relation
domain = (10, 1, 7, 1, 1)
# Vectorize relation
x = transformation.Vectorize('CPS', reduced_domain=domain).transform(R)
# Setup arbitrary constants for MWEM
seed = 0
ratio = 0.5
rounds = 3
data_scale = 1e5
use_history = True
epsilon = 0.1
# Create query workload
W = workload.RandomRange(None, (np.prod(domain), ), 25)
# Calculate noisy estimate of x
x_hat = standalone.Mwem(ratio, rounds, data_scale, domain,
use_history).Run(W, x, epsilon, seed)
class TestPlans(unittest.TestCase):
seed = 10
relation_cps = RelationHelper('CPS').load()
relation_stroke = RelationHelper('STROKE_2D').load()
cps_domain = (10, 2, 7, 2, 2)
stroke_domain = (64, 64)
x_cps = transformation.Vectorize('CPS', reduced_domain=cps_domain).transform(relation_cps)
x_cps_scale = x_cps.sum()
x_stroke = transformation.Vectorize('STROKE', reduced_domain=stroke_domain).transform(relation_stroke)
W_cps = workload.RandomRange(None, (len(x_cps),), 25)
W_stroke = workload.RandomRange(None, (len(x_stroke),), 25)
def setUp(self):
self.eps = 0.1
def test_identity(self):
x_hat = standalone.Identity().Run(self.W_cps,
self.x_cps,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_privelet(self):
x_hat = standalone.Privelet().Run(self.W_cps,
self.x_cps,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_h2(self):
x_hat = standalone.H2().Run(self.W_cps,
self.x_cps,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_hb(self):
domain_shape = (len(self.x_cps),)
x_hat = standalone.HB(domain_shape).Run(self.W_cps,
self.x_cps,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_hb_2D(self):
x_hat = standalone.HB(self.stroke_domain).Run(self.W_stroke,
self.x_stroke,
self.eps,
self.seed)
self.W_stroke.get_matrix() * x_hat
def test_greedy_h(self):
x_hat = standalone.GreedyH().Run(self.W_cps,
self.x_cps,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_uniform(self):
x_hat = standalone.Uniform().Run(self.W_cps,
self.x_cps,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_privBayesLS(self):
theta = 1
x_hat = standalone.PrivBayesLS(theta, self.cps_domain).Run(self.W_cps,
self.relation_cps,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_mwem(self):
ratio = 0.5
rounds = 3
data_scale = 1e5
domain_shape = (len(self.x_cps),)
use_history = True
x_hat = standalone.Mwem(ratio,
rounds,
data_scale,
domain_shape,
use_history).Run(self.W_cps,
self.x_cps,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_mwem_2D(self):
ratio = 0.5
rounds = 3
data_scale = 1e5
use_history = True
x_hat = standalone.Mwem(ratio,
rounds,
data_scale,
self.stroke_domain,
use_history).Run(self.W_stroke,
self.x_stroke,
self.eps,
self.seed)
self.W_stroke.get_matrix() * x_hat
def test_ahp(self):
eta = 0.35
ratio = 0.85
x_hat = standalone.Ahp(eta, ratio).Run(self.W_cps,
self.x_cps,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_dawa(self):
ratio = 0.25
approx = False
domain_shape = (len(self.x_cps),)
x_hat = standalone.Dawa(domain_shape, ratio, approx).Run(self.W_cps,
self.x_cps,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_dawa_2D(self):
ratio = 0.25
approx = False
x_hat = standalone.Dawa(self.stroke_domain, ratio, approx).Run(self.W_stroke,
self.x_stroke,
self.eps,
self.seed)
self.W_stroke.get_matrix() * x_hat
def test_quad_tree(self):
x_hat = standalone.QuadTree().Run(self.W_cps,
self.x_cps,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_ugrid(self):
data_scale = 1e5
x = self.x_cps.reshape((len(self.x_cps) // 2, 2))
x_hat = standalone.UGrid(data_scale).Run(self.W_cps,
x,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_agrid(self):
data_scale = 1e5
x = self.x_cps.reshape((len(self.x_cps) // 2, 2))
x_hat = standalone.AGrid(data_scale).Run(self.W_cps,
x,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_dawa_striped(self):
stripe_dim = 0
ratio = 0.25
approx = False
x_hat = standalone.DawaStriped(ratio, self.cps_domain, stripe_dim, approx).Run(self.W_cps,
self.x_cps,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_striped_HB_slow(self):
stripe_dim = 0
x_hat = standalone.StripedHB(self.cps_domain, stripe_dim).Run(self.W_cps,
self.x_cps,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_mwem_variant_b(self):
ratio = 0.5
rounds = 3
x_hat = standalone.MwemVariantB(ratio, rounds, self.x_cps_scale, self.cps_domain, True).Run(self.W_cps,
self.x_cps,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_mwem_variant_c(self):
ratio = 0.5
rounds = 3
x_hat = standalone.MwemVariantC(ratio, rounds, self.x_cps_scale, self.cps_domain, True).Run(self.W_cps,
self.x_cps,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat
def test_mwem_variant_d(self):
ratio = 0.5
rounds = 3
x_hat = standalone.MwemVariantD(ratio, rounds, self.x_cps_scale, self.cps_domain, True).Run(self.W_cps,
self.x_cps,
self.eps,
self.seed)
self.W_cps.get_matrix() * x_hat