def Test_sampling_error(self):
"""
Testing :meth:`bet.calculateP.calculateError.sampling_error`
"""
num = self.disc.check_nums()
neiList = calculateError.cell_connectivity_exact(self.disc)
for i in range(num):
self.assertGreater(len(neiList[i]), 0)
(B_N, C_N) = calculateError.boundary_sets(self.disc, neiList)
s_error = calculateError.sampling_error(self.disc, exact=True)
(upper, lower) = s_error.calculate_for_contour_events()
for x in upper:
if not np.isnan(x):
self.assertGreaterEqual(x, 0.0)
for x in lower:
if not np.isnan(x):
self.assertLessEqual(x, 0.0)
s_set = self.disc._input_sample_set.copy()
regions_local = np.equal(self.disc._io_ptr_local, 0)
s_set.set_region_local(regions_local)
s_set.local_to_global()
(up, low) = s_error.calculate_for_sample_set_region(s_set,
1)
if not np.isnan(up):
self.assertAlmostEqual(up, upper[0])
else:
self.assertTrue(np.isnan(upper[0]))
if not np.isnan(low):
self.assertAlmostEqual(low, lower[0])
else:
self.assertTrue(np.isnan(lower[0]))
(up, low) = s_error.calculate_for_sample_set_region(s_set,
1,
emulated_set=self.disc._input_sample_set)
if np.isnan(up):
self.assertEqual(np.isnan(upper[0]), True)
else:
self.assertAlmostEqual(up, upper[0])
if np.isnan(low):
self.assertEqual(np.isnan(lower[0]), True)
else:
self.assertAlmostEqual(low, lower[0])
self.disc.set_emulated_input_sample_set(self.disc._input_sample_set)
(up, low) = s_error.calculate_for_sample_set_region(s_set,
1)
if np.isnan(up):
self.assertEqual(np.isnan(upper[0]), True)
else:
self.assertAlmostEqual(up, upper[0])
if np.isnan(low):
self.assertEqual(np.isnan(lower[0]), True)
else:
self.assertAlmostEqual(low, lower[0])
def Test_sampling_error(self):
"""
Testing :meth:`bet.calculateP.calculateError.sampling_error`
"""
num = self.disc.check_nums()
neiList = calculateError.cell_connectivity_exact(self.disc)
for i in range(num):
self.assertGreater(len(neiList[i]), 0)
(B_N, C_N) = calculateError.boundary_sets(self.disc, neiList)
s_error = calculateError.sampling_error(self.disc, exact=True)
(upper, lower) = s_error.calculate_for_contour_events()
for x in upper:
if not np.isnan(x):
self.assertGreaterEqual(x, 0.0)
for x in lower:
if not np.isnan(x):
self.assertLessEqual(x, 0.0)
s_set = self.disc._input_sample_set.copy()
regions_local = np.equal(self.disc._io_ptr_local, 0)
s_set.set_region_local(regions_local)
s_set.local_to_global()
(up, low) = s_error.calculate_for_sample_set_region(s_set, 1)
if not np.isnan(up):
self.assertAlmostEqual(up, upper[0])
else:
self.assertTrue(np.isnan(upper[0]))
if not np.isnan(low):
self.assertAlmostEqual(low, lower[0])
else:
self.assertTrue(np.isnan(lower[0]))
(up, low) = s_error.calculate_for_sample_set_region(
s_set, 1, emulated_set=self.disc._input_sample_set)
if np.isnan(up):
self.assertEqual(isnan(upper[0]), True)
else:
self.assertAlmostEqual(up, upper[0])
if np.isnan(low):
self.assertEqual(isnan(lower[0]), True)
else:
self.assertAlmostEqual(low, lower[0])
self.disc.set_emulated_input_sample_set(self.disc._input_sample_set)
(up, low) = s_error.calculate_for_sample_set_region(s_set, 1)
if np.isnan(up):
self.assertEqual(isnan(upper[0]), True)
else:
self.assertAlmostEqual(up, upper[0])
if np.isnan(low):
self.assertEqual(isnan(lower[0]), True)
else:
self.assertAlmostEqual(low, lower[0])
sur = surrogates.piecewise_polynomial_surrogate(my_disc_exact)
sur_disc_lin = sur.generate_for_input_set(emulated_inputs, order=1)
(Pt, _) = sur.calculate_prob_for_sample_set_region(s_set,
regions=[0])
P_true = Pt[0]
if comm.rank == 0:
print "True probability: ", P_true
# Approximate probabilty
P_approx = calculateP.prob_from_sample_set(my_disc._input_sample_set,
s_set)[0]
if comm.rank == 0:
print "Approximate probability: ", P_approx
# Calculate sampling error
samp_er = calculateError.sampling_error(my_disc)
(up, low) = samp_er.calculate_for_sample_set_region(s_set,
region=0, emulated_set=emulated_inputs)
if comm.rank == 0:
print "Sampling error upper bound: ", up
print "Sampling error lower bound: ", low
# Calculate modeling error
mod_er = calculateError.model_error(my_disc)
er_est = mod_er.calculate_for_sample_set_region(s_set,
region=0, emulated_set=emulated_inputs)
if comm.rank == 0:
print "Modeling error estimate: ", er_est
# Make piecewise constant surrogate of discretization
# Calculate true probabilty using linear surrogate on true data
sur = surrogates.piecewise_polynomial_surrogate(my_disc_exact)
sur_disc_lin = sur.generate_for_input_set(emulated_inputs, order=1)
(Pt, _) = sur.calculate_prob_for_sample_set_region(s_set, regions=[0])
P_true = Pt[0]
if comm.rank == 0:
print "True probability: ", P_true
# Approximate probabilty
P_approx = calculateP.prob_from_sample_set(my_disc._input_sample_set, s_set)[0]
if comm.rank == 0:
print "Approximate probability: ", P_approx
# Calculate sampling error
samp_er = calculateError.sampling_error(my_disc)
(up,
low) = samp_er.calculate_for_sample_set_region(s_set,
region=0,
emulated_set=emulated_inputs)
if comm.rank == 0:
print "Sampling error upper bound: ", up
print "Sampling error lower bound: ", low
# Calculate modeling error
mod_er = calculateError.model_error(my_disc)
er_est = mod_er.calculate_for_sample_set_region(s_set,
region=0,
emulated_set=emulated_inputs)
if comm.rank == 0:
print "Modeling error estimate: ", er_est