def test_fit_pytorchgan_continuous_no_transfromer(self):
dpgan = PytorchDPSynthesizer(
eps, PATECTGAN(epsilon=eps, batch_size=batch_size), None)
pd_data_xy = pd.DataFrame(np_data_xy, columns=["x", "y"])
try:
dpgan.fit(pd_data_xy, categorical_columns=['x'])
except ValueError:
return
raise AssertionError('PATECTGAN should have raised a ValueError')
class TestPytorchDPSynthesizer_PATECTGAN:
def setup(self):
self.patectgan = PytorchDPSynthesizer(1.0, PATECTGAN(), None)
def test_fit(self):
self.patectgan.fit(
df_non_continuous,
categorical_columns=['sex', 'educ', 'race', 'married'])
assert self.patectgan.gan._generator
def test_sample(self):
self.patectgan.fit(
df_non_continuous,
categorical_columns=['sex', 'educ', 'race', 'married'])
sample_size = len(df)
synth_data = self.patectgan.sample(sample_size)
assert synth_data.shape == df_non_continuous.shape
class TestDPGAN:
def setup(self):
self.pategan = PytorchDPSynthesizer(1.0, PATEGAN(1.0),
GeneralTransformer())
def test_fit(self):
df_non_continuous = df[['sex', 'educ', 'race', 'married']]
self.pategan.fit(
df_non_continuous,
categorical_columns=['sex', 'educ', 'race', 'married'])
assert self.pategan.gan.generator
def test_sample(self):
df_non_continuous = df[['sex', 'educ', 'race', 'married']]
self.pategan.fit(
df_non_continuous,
categorical_columns=['sex', 'educ', 'race', 'married'])
sample_size = len(df_non_continuous)
synth_data = self.pategan.sample(sample_size)
assert synth_data.shape == df_non_continuous.shape
class TestPytorchDPSynthesizer_DPCTGAN:
def setup(self):
self.dpctgan = PytorchDPSynthesizer(1.0, DPCTGAN(), None)
def test_fit(self):
df_non_continuous = df[['sex', 'educ', 'race', 'married']]
self.dpctgan.fit(
df_non_continuous,
categorical_columns=['sex', 'educ', 'race', 'married'])
assert self.dpctgan.gan._generator
def test_sample(self):
self.dpctgan.fit(
df_non_continuous,
categorical_columns=['sex', 'educ', 'race', 'married'])
sample_size = len(df)
synth_data = self.dpctgan.sample(sample_size)
assert synth_data.shape == df_non_continuous.shape
def test_fit_numpy(self):
dpctgan = DPCTGAN(epsilon=1.0)
dpctgan.train(nf_non_continuous, categorical_columns=[0, 1, 2, 3])
class TestPytorchDPSynthesizer_DPGAN:
def setup(self):
self.dpgan = PytorchDPSynthesizer(1.0, DPGAN(), GeneralTransformer())
def test_fit(self):
self.dpgan.fit(df_non_continuous,
categorical_columns=['sex', 'educ', 'race', 'married'])
assert self.dpgan.gan.generator
def test_sample(self):
self.dpgan.fit(df_non_continuous,
categorical_columns=['sex', 'educ', 'race', 'married'])
sample_size = len(df_non_continuous)
synth_data = self.dpgan.sample(sample_size)
assert synth_data.shape == df_non_continuous.shape
def test_fit_continuous(self):
dpgan = DPGAN(epsilon=1.0)
df_continuous = df[['age', 'educ', 'income']]
dpgan.train(df_continuous)
synth_data = dpgan.generate(len(df_continuous))
assert synth_data.shape == df_continuous.shape
def QuailSynth(epsilon):
return PytorchDPSynthesizer(epsilon=epsilon, preprocessor=None,
gan=PATECTGAN(loss='cross_entropy', batch_size=50, pac=1))
def setup(self):
self.pategan = PytorchDPSynthesizer(1.0, PATEGAN(1.0),
GeneralTransformer())
def setup(self):
self.patectgan = PytorchDPSynthesizer(
1.0, PATECTGAN(regularization='dragan'), None)
def setup(self):
self.patectgan = PytorchDPSynthesizer(1.0, PATECTGAN(), None)
def setup(self):
self.dpctgan = PytorchDPSynthesizer(1.0, DPCTGAN(), None)
def setup(self):
self.dpgan = PytorchDPSynthesizer(1.0, DPGAN(), GeneralTransformer())