Fix LGTM alerts

Signed-off-by: Beat Buesser <[email protected]>

Author: Beat Buesser

art/attacks/evasion/hop_skip_jump.py

@@ -24,7 +24,7 @@
 from __future__ import absolute_import, division, print_function, unicode_literals
 
 import logging
-from typing import Optional, Tuple, Union, List, TYPE_CHECKING
+from typing import Optional, Tuple, Union, TYPE_CHECKING
 
 import numpy as np
 from tqdm import tqdm

art/attacks/evasion/imperceptible_asr/imperceptible_asr.py

@@ -194,7 +194,6 @@ def _create_adversarial(self, x, y) -> np.ndarray:
         x_adversarial = [None] * batch_size
 
         x_perturbed = x.copy()
-        perturbation = np.zeros_like(x_perturbed)
 
         for i in range(self.max_iter_1):
             # perform FGSM step for x

art/estimators/certification/randomized_smoothing/numpy.py

@@ -24,6 +24,7 @@
 
 import numpy as np
 
+from art.estimators.estimator import BaseEstimator
 from art.estimators.certification.randomized_smoothing.randomized_smoothing import RandomizedSmoothingMixin
 from art.estimators.classification import ClassifierMixin, ClassGradientsMixin
 import logging
@@ -35,7 +36,7 @@
 logger = logging.getLogger(__name__)
 
 
-class NumpyRandomizedSmoothing(RandomizedSmoothingMixin, ClassGradientsMixin, ClassifierMixin):
+class NumpyRandomizedSmoothing(RandomizedSmoothingMixin, ClassGradientsMixin, ClassifierMixin, BaseEstimator):
     """
     Implementation of Randomized Smoothing applied to classifier predictions and gradients, as introduced
     in Cohen et al. (2019).
@@ -53,12 +54,20 @@ def __init__(
         :param scale: Standard deviation of Gaussian noise added.
         :param alpha: The failure probability of smoothing
         """
-        super().__init__(classifier)
+        super().__init__(
+            model=None,
+            input_shape=classifier.input_shape,
+            nb_classes=classifier.nb_classes,
+            channels_first=classifier.channels_first,
+            clip_values=classifier.clip_values,
+            preprocessing_defences=classifier.preprocessing_defences,
+            postprocessing_defences=classifier.postprocessing_defences,
+            preprocessing=classifier.preprocessing,
+            sample_size=sample_size,
+            scale=scale,
+            alpha=alpha,
+        )
         self.classifier = classifier
-        self._nb_classes = classifier.nb_classes
-        self.sample_size = sample_size
-        self.scale = scale
-        self.alpha = alpha
 
     def _predict_classifier(self, x: np.ndarray, batch_size: int) -> np.ndarray:
         """

conftest.py

@@ -51,8 +51,6 @@
 def get_default_framework():
     import tensorflow as tf
 
-    default_framework = "tensorflow"
-
     if tf.__version__[0] == "2":
         default_framework = "tensorflow2"
     else:
@@ -343,7 +341,7 @@ def _expected_values():
 
 @pytest.fixture(scope="session")
 def get_image_classifier_mx_model():
-    import mxnet
+    import mxnet  # lgtm [py/import-and-import-from]
 
     # TODO needs to be made parameterizable once Mxnet allows multiple identical models to be created in one session
     from_logits = True
@@ -374,7 +372,7 @@ def forward(self, x):
 
 @pytest.fixture
 def get_image_classifier_mx_instance(get_image_classifier_mx_model, mnist_shape):
-    import mxnet
+    import mxnet  # lgtm [py/import-and-import-from]
     from art.estimators.classification import MXClassifier
 
     model = get_image_classifier_mx_model