9b825ea2437603451cbbdfe07596b820f4523f36,src/metrics_unittest.py,TestMinimalPerturbations,test_mnist,#TestMinimalPerturbations#,45

Before Change

        // get classifier
        im_shape = X_train[0].shape
        model = cnn.cnn_model(im_shape, act="relu")
        model.compile(loss="categorical_crossentropy", optimizer="adam", metrics=["accuracy"])

        // Fit the model
        model.fit(X_train, Y_train, epochs=1, batch_size=BATCH_SIZE)

After Change

        session = tf.Session()
        K.set_session(session)

        comp_params = {"loss": "categorical_crossentropy",
                       "optimizer": "adam",
                       "metrics": ["accuracy"]}

        // get MNIST
        (X_train, Y_train), (_, _) = load_mnist()
        X_train, Y_train = X_train[:NB_TRAIN], Y_train[:NB_TRAIN]

        // get classifier
        im_shape = X_train[0].shape
        classifier = CNN(im_shape, act="relu")
        classifier.compile(comp_params)

        // Fit the classifier
        classifier.fit(X_train, Y_train, epochs=1, batch_size=BATCH_SIZE)

        // Compute minimal perturbations
        params = {"eps_step":1.1,
                  "clip_min":0.,
                  "clip_max":1.}

        emp_robust = empirical_robustness(X_train, classifier.model, session, "fgsm", params)
        self.assertEqual(emp_robust, 0.)

        params = {"eps_step": 1.,
                  "eps_max": 1.,
                  "clip_min": None,
                  "clip_max": None}
        emp_robust = empirical_robustness(X_train, classifier.model, session, "fgsm", params)
        self.assertAlmostEqual(emp_robust*LA.norm(X_train), 1., emp_robust**LA.norm(X_train))

        // params = {"theta": 1.,

Instances