def test_regularized_logstic_cost_function_val(self):
x_train = TEST_DATA['rcost']['x_train']
y_train = TEST_DATA['rcost']['y_train']
w = TEST_DATA['rcost']['w']
reg_lambda = TEST_DATA['rcost']['lambda']
val = TEST_DATA['rcost']['L']
val_computed, _ = regularized_logistic_cost_function(
w, x_train, y_train, reg_lambda)
self.assertAlmostEqual(val, val_computed, 6)
def test_regularized_logstic_cost_function_grad(self):
x_train = TEST_DATA['rcost']['x_train']
y_train = TEST_DATA['rcost']['y_train']
w = TEST_DATA['rcost']['w']
reg_lambda = TEST_DATA['rcost']['lambda']
grad = TEST_DATA['rcost']['grad']
_, grad_computed = regularized_logistic_cost_function(
w, x_train, y_train, reg_lambda)
max_diff = np.max(np.abs(grad - grad_computed))
self.assertAlmostEqual(max_diff, 0, 6)
def test_regularized_logstic_cost_function_grad(self):
w = TEST_DATA['rcost']['w']
x_train = TEST_DATA['rcost']['x_train']
y_train = TEST_DATA['rcost']['y_train']
reg_lambda = TEST_DATA['rcost']['lambda']
grad_expected = TEST_DATA['rcost']['grad']
_, grad = regularized_logistic_cost_function(w, x_train, y_train, reg_lambda)
self.assertEqual(np.shape(grad), (20, 1))
np.testing.assert_almost_equal(grad, grad_expected)
def test_regularized_logstic_cost_function_val(self):
w = TEST_DATA['rcost']['w']
x_train = TEST_DATA['rcost']['x_train']
y_train = TEST_DATA['rcost']['y_train']
reg_lambda = TEST_DATA['rcost']['lambda']
val_expected = TEST_DATA['rcost']['L']
val, _ = regularized_logistic_cost_function(w, x_train, y_train, reg_lambda)
self.assertEqual(np.size(val), 1)
self.assertAlmostEqual(val, val_expected)