def setUp(self):
dataset_filepath = SurealConfig.test_resource_path('NFLX_dataset_public_raw.py')
dataset = import_python_file(dataset_filepath)
np.random.seed(0)
info_dict = {
'corrupt_probability': 0.1,
}
self.dataset_reader = CorruptDataRawDatasetReader(dataset, input_dict=info_dict)
class CorruptDataDatasetReaderTest(unittest.TestCase):
def setUp(self):
dataset_filepath = SurealConfig.test_resource_path(
'NFLX_dataset_public_raw.py')
dataset = import_python_file(dataset_filepath)
np.random.seed(0)
info_dict = {
'corrupt_probability': 0.1,
}
self.dataset_reader = CorruptDataRawDatasetReader(dataset,
input_dict=info_dict)
def test_opinion_score_2darray(self):
os_2darray = self.dataset_reader.opinion_score_2darray
self.assertAlmostEqual(np.mean(np.std(os_2darray, axis=1)),
0.79796204942957094,
places=4)
def test_to_dataset(self):
dataset = self.dataset_reader.to_dataset()
old_scores = [
dis_video['os']
for dis_video in self.dataset_reader.dataset.dis_videos
]
new_scores = [dis_video['os'] for dis_video in dataset.dis_videos]
self.assertNotEqual(old_scores, new_scores)
def run_one_corrput_prob(corrupt_prob, dataset, seed):
np.random.seed(seed)
info_dict = {
'corrupt_probability': corrupt_prob,
}
dataset_reader = CorruptDataRawDatasetReader(dataset,
input_dict=info_dict)
subjective_model = model_class(dataset_reader)
try:
result = subjective_model.run_modeling(normalize_final=False)
except ValueError as e:
print 'Warning: {}, return result None'.format(e)
result = None
return dataset_reader, result