def setUp(self):
dataset_filepath = SurealConfig.test_resource_path(
'NFLX_dataset_public_raw.py')
dataset = import_python_file(dataset_filepath)
dataset_reader = RawDatasetReader(dataset)
pc_dataset = dataset_reader.to_pc_dataset(pc_type='within_subject')
self.pc_dataset_reader = PairedCompDatasetReader(pc_dataset)
def test_dataset_to_pc_dataset(self):
pc_dataset = self.dataset_reader.to_pc_dataset()
pc_dataset_reader = PairedCompDatasetReader(pc_dataset)
opinion_score_3darray = pc_dataset_reader.opinion_score_3darray
self.assertEqual(np.nansum(opinion_score_3darray), 8242)
self.assertEqual(np.nanmean(opinion_score_3darray), 0.816039603960396)
self.assertEqual(np.nanmin(opinion_score_3darray), 0.5)
self.assertEqual(np.nanmax(opinion_score_3darray), 1.0)
def from_dataset_file(cls,
dataset_filepath,
content_ids=None,
asset_ids=None):
dataset = cls._import_dataset_and_filter(dataset_filepath, content_ids,
asset_ids)
dataset_reader = PairedCompDatasetReader(dataset)
return cls(dataset_reader)
def test_dataset_to_pc_dataset_within_subject_per_asset_mean_scores(self):
pc_dataset = self.dataset_reader.to_pc_dataset(pc_type='within_subject', per_asset_mean_scores=np.ones(79))
pc_dataset_reader = PairedCompDatasetReader(pc_dataset)
opinion_score_3darray = pc_dataset_reader.opinion_score_3darray
self.assertEqual(np.nansum(opinion_score_3darray), 80106)
self.assertEqual(np.nanmean(opinion_score_3darray), 0.5)
self.assertEqual(np.nanmin(opinion_score_3darray), 0.5)
self.assertEqual(np.nanmax(opinion_score_3darray), 0.5)
def test_dataset_to_pc_dataset_coin_toss(self):
pc_dataset = self.dataset_reader.to_pc_dataset(tiebreak_method='coin_toss')
pc_dataset_reader = PairedCompDatasetReader(pc_dataset)
opinion_score_3darray = pc_dataset_reader.opinion_score_3darray
self.assertEqual(np.nansum(opinion_score_3darray), 8242)
self.assertEqual(np.nanmean(opinion_score_3darray), 1.0)
self.assertEqual(np.nanmin(opinion_score_3darray), 1.0)
self.assertEqual(np.nanmax(opinion_score_3darray), 1.0)
def test_dataset_to_pc_dataset_within_subject(self):
pc_dataset = self.dataset_reader.to_pc_dataset(pc_type='within_subject')
pc_dataset_reader = PairedCompDatasetReader(pc_dataset)
opinion_score_3darray = pc_dataset_reader.opinion_score_3darray
self.assertEqual(np.nansum(opinion_score_3darray), 80106)
self.assertEqual(np.nanmean(opinion_score_3darray), 0.8050935185278244)
self.assertEqual(np.nanmin(opinion_score_3darray), 0.5)
self.assertEqual(np.nanmax(opinion_score_3darray), 1.0)
def test_dataset_to_pc_dataset_sampling_rate(self):
import random
random.seed(0)
pc_dataset = self.dataset_reader.to_pc_dataset(sampling_rate=0.1)
pc_dataset_reader = PairedCompDatasetReader(pc_dataset)
opinion_score_3darray = pc_dataset_reader.opinion_score_3darray
self.assertEqual(np.nansum(opinion_score_3darray), 844)
self.assertEqual(np.nanmean(opinion_score_3darray), 0.8045757864632984)
self.assertEqual(np.nanmin(opinion_score_3darray), 0.5)
self.assertEqual(np.nanmax(opinion_score_3darray), 1.0)
def test_dataset_to_pc_dataset_random(self):
import random
random.seed(0)
pc_dataset = self.dataset_reader.to_pc_dataset(randomness_level=0.5)
pc_dataset_reader = PairedCompDatasetReader(pc_dataset)
opinion_score_3darray = pc_dataset_reader.opinion_score_3darray
self.assertEqual(np.nansum(opinion_score_3darray), 8242)
self.assertEqual(np.nanmean(opinion_score_3darray), 0.8973326075122482)
self.assertEqual(np.nanmin(opinion_score_3darray), 0.5)
self.assertEqual(np.nanmax(opinion_score_3darray), 1.0)
def test_dataset_to_pc_dataset_per_asset_noise_levels(self):
import random
random.seed(0)
pc_dataset = self.dataset_reader.to_pc_dataset(per_asset_noise_levels=np.hstack([np.ones(39), np.ones(40) * 0.1]))
pc_dataset_reader = PairedCompDatasetReader(pc_dataset)
opinion_score_3darray = pc_dataset_reader.opinion_score_3darray
self.assertEqual(np.nansum(opinion_score_3darray), 8242)
self.assertAlmostEqual(float(np.nanmean(opinion_score_3darray)), 1.0, delta=0.0001)
self.assertEqual(np.nanmin(opinion_score_3darray), 1.0)
self.assertEqual(np.nanmax(opinion_score_3darray), 1.0)
def test_dataset_to_pc_dataset_per_asset_sampling_rates(self):
import random
random.seed(0)
pc_dataset = self.dataset_reader.to_pc_dataset(per_asset_sampling_rates=np.hstack([np.ones(39), np.ones(40) * 0.1]))
pc_dataset_reader = PairedCompDatasetReader(pc_dataset)
opinion_score_3darray = pc_dataset_reader.opinion_score_3darray
self.assertEqual(np.nansum(opinion_score_3darray), 4546)
self.assertAlmostEqual(float(np.nanmean(opinion_score_3darray)), 0.8116303960042811, delta=0.01)
self.assertEqual(np.nanmin(opinion_score_3darray), 0.5)
self.assertEqual(np.nanmax(opinion_score_3darray), 1.0)
def test_dataset_to_pc_dataset_sampling_rate_greater_than_1(self):
import random
random.seed(0)
pc_dataset = self.dataset_reader.to_pc_dataset(sampling_rate=2.1)
pc_dataset_reader = PairedCompDatasetReader(pc_dataset)
opinion_score_3darray = pc_dataset_reader.opinion_score_3darray
self.assertEqual(np.nansum(opinion_score_3darray), 8242)
self.assertAlmostEqual(float(np.nanmean(opinion_score_3darray)), 0.816039603960396, delta=0.0001)
self.assertEqual(np.nanmin(opinion_score_3darray), 0.5)
self.assertEqual(np.nanmax(opinion_score_3darray), 1.0)
def test_dataset_to_pc_dataset_random(self):
import random
random.seed(0)
pc_dataset = self.dataset_reader.to_pc_dataset(cointoss_rate=0.5)
pc_dataset_reader = PairedCompDatasetReader(pc_dataset)
opinion_score_3darray = pc_dataset_reader.opinion_score_3darray
self.assertEqual(np.nansum(opinion_score_3darray), 8242)
# check: python2 values seem to fluctuate quite a bit
self.assertAlmostEqual(float(np.nanmean(opinion_score_3darray)), 0.8966492602262838, delta=0.01)
self.assertEqual(np.nanmin(opinion_score_3darray), 0.5)
self.assertEqual(np.nanmax(opinion_score_3darray), 1.0)
def test_dataset_to_pc_dataset_per_asset_sampling_rates(self):
import random
random.seed(0)
pc_dataset = self.dataset_reader.to_pc_dataset(
per_asset_sampling_rates=np.hstack(
[np.ones(39), np.ones(40) * 0.1]))
pc_dataset_reader = PairedCompDatasetReader(pc_dataset)
opinion_score_3darray = pc_dataset_reader.opinion_score_3darray
self.assertEqual(np.nansum(opinion_score_3darray), 4550)
self.assertEqual(np.nanmean(opinion_score_3darray), 0.8098967604129583)
self.assertEqual(np.nanmin(opinion_score_3darray), 0.5)
self.assertEqual(np.nanmax(opinion_score_3darray), 1.0)
def test_dataset_to_pc_dataset_sampling_rate(self):
import random
random.seed(0)
pc_dataset = self.dataset_reader.to_pc_dataset(sampling_rate=0.1)
pc_dataset_reader = PairedCompDatasetReader(pc_dataset)
opinion_score_3darray = pc_dataset_reader.opinion_score_3darray
self.assertEqual(np.nansum(opinion_score_3darray), 844)
self.assertAlmostEqual(float(np.nanmean(opinion_score_3darray)),
0.8107588856868396,
delta=0.1)
self.assertEqual(np.nanmin(opinion_score_3darray), 0.5)
self.assertEqual(np.nanmax(opinion_score_3darray), 1.0)
def visualize_pc_dataset(dataset_filepath):
dataset = import_python_file(dataset_filepath)
dataset_reader = PairedCompDatasetReader(dataset)
tensor_pvs_pvs_subject = dataset_reader.opinion_score_3darray
plt.figure()
# plot the rate of winning x, 0 <= x <= 1.0, of one PVS compared against another PVS
mtx_pvs_pvs = np.nansum(tensor_pvs_pvs_subject, axis=2) \
/ (np.nansum(tensor_pvs_pvs_subject, axis=2) +
np.nansum(tensor_pvs_pvs_subject, axis=2).transpose())
plt.imshow(mtx_pvs_pvs, interpolation='nearest')
plt.title(r'Paired Comparison Winning Rate')
plt.ylabel(r"PVS ($j$)")
plt.xlabel(r"PVS ($j'$) [Compared Against]")
plt.set_cmap('jet')
plt.colorbar()
plt.tight_layout()
def setUp(self):
pc_dataset = import_python_file(
SurealConfig.test_resource_path('lukas_pc_dataset.py'))
self.pc_dataset_reader = PairedCompDatasetReader(pc_dataset)