def __init__(self):
dataset = data.german_credit_numeric()
del dataset['test_features']
del dataset['test_labels']
super(GermanCreditNumericLogisticRegression, self).__init__(
name='german_credit_numeric_logistic_regression',
pretty_name='German Credit Numeric Logistic Regression',
**dataset)
def load_dataset_in_xla():
dataset = data.german_credit_numeric()
# The actual dataset loading will happen in Eager mode, courtesy of the
# init_scope.
return (
tf.convert_to_tensor(dataset['train_features']),
tf.convert_to_tensor(dataset['train_labels']),
)
def german_credit_numeric_logistic_regression():
"""German credit (numeric) logistic regression.
Returns:
target: StanModel.
"""
dataset = data.german_credit_numeric()
del dataset['test_features']
del dataset['test_labels']
return logistic_regression.logistic_regression(**dataset)
def german_credit_numeric_sparse_logistic_regression():
"""German credit (numeric) logistic regression with a sparsity-inducing prior.
Returns:
target: StanModel.
"""
dataset = data.german_credit_numeric()
del dataset['test_features']
del dataset['test_labels']
return sparse_logistic_regression.sparse_logistic_regression(**dataset)
def testGermanCreditNumeric(self):
dataset = data.german_credit_numeric(train_fraction=0.75)
self.assertEqual((750, 24), dataset['train_features'].shape)
self.assertEqual((750,), dataset['train_labels'].shape)
self.assertEqual((250, 24), dataset['test_features'].shape)
self.assertEqual((250,), dataset['test_labels'].shape)
self.assertAllClose(
np.zeros([24]), dataset['train_features'].mean(0), atol=1e-5)
self.assertAllClose(
np.ones([24]), dataset['train_features'].std(0), atol=1e-5)
self.assertAllClose(
np.zeros([24]), dataset['test_features'].mean(0), atol=0.3)
self.assertAllClose(
np.ones([24]), dataset['test_features'].std(0), atol=0.3)