def test_emp_robustness_mnist(self):
(x_train, y_train), (_, _), _, _ = load_mnist()
x_train, y_train = x_train[:NB_TRAIN], y_train[:NB_TRAIN]
# Get classifier
classifier = self._cnn_mnist_k([28, 28, 1])
classifier.fit(x_train,
y_train,
batch_size=BATCH_SIZE,
nb_epochs=2,
verbose=0)
# Compute minimal perturbations
params = {"eps_step": 1.1}
emp_robust = empirical_robustness(classifier, x_train, str("fgsm"),
params)
self.assertEqual(emp_robust, 0.0)
params = {"eps_step": 1.0, "eps": 1.0}
emp_robust = empirical_robustness(classifier, x_train, str("fgsm"),
params)
self.assertAlmostEqual(emp_robust, 1.000369094488189, 4)
params = {"eps_step": 0.1, "eps": 0.2}
emp_robust = empirical_robustness(classifier, x_train, str("fgsm"),
params)
self.assertLessEqual(emp_robust, 0.65)
def test_emp_robustness_mnist(self):
(x_train, y_train), (_, _), _, _ = load_mnist()
x_train, y_train = x_train[:NB_TRAIN], y_train[:NB_TRAIN]
# Get classifier
classifier = self._cnn_mnist_k([28, 28, 1])
classifier.fit(x_train, y_train, batch_size=BATCH_SIZE, nb_epochs=2)
# Compute minimal perturbations
params = {"eps_step": 1.1}
emp_robust = empirical_robustness(classifier, x_train, str('fgsm'),
params)
self.assertEqual(emp_robust, 0.)
params = {"eps_step": 1., "eps": 1.}
emp_robust = empirical_robustness(classifier, x_train, str('fgsm'),
params)
self.assertAlmostEqual(emp_robust, 0.5006149157681419, 3)
params = {"eps_step": 0.1, "eps": 0.2}
emp_robust = empirical_robustness(classifier, x_train, str('fgsm'),
params)
self.assertLessEqual(emp_robust, 0.65)