Merge branch 'dev_1.8.0' into development_issue_1196

Author: beat-buesser

art/classifiers/__init__.py

@@ -16,6 +16,7 @@
 from art.estimators.classification.lightgbm import LightGBMClassifier
 from art.estimators.classification.mxnet import MXClassifier
 from art.estimators.classification.pytorch import PyTorchClassifier
+from art.estimators.classification.query_efficient_bb import QueryEfficientGradientEstimationClassifier
 from art.estimators.classification.scikitlearn import SklearnClassifier
 from art.estimators.classification.tensorflow import (
     TFClassifier,

art/classifiers/scikitlearn/__init__.py

@@ -39,7 +39,7 @@
 
 class BlackBoxClassifier(ClassifierMixin, BaseEstimator):
     """
-    Wrapper class for black-box classifiers.
+    Class for black-box classifiers.
     """
 
     estimator_params = Classifier.estimator_params + ["nb_classes", "input_shape", "predict_fn"]
@@ -163,7 +163,7 @@ def save(self, filename: str, path: Optional[str] = None) -> None:
 
 class BlackBoxClassifierNeuralNetwork(NeuralNetworkMixin, ClassifierMixin, BaseEstimator):
     """
-    Wrapper class for black-box neural network classifiers.
+    Class for black-box neural network classifiers.
     """
 
     estimator_params = (

art/estimators/classification/__init__.py

@@ -43,7 +43,7 @@
 
 class CatBoostARTClassifier(ClassifierDecisionTree):
     """
-    Wrapper class for importing CatBoost models.
+    Class for importing CatBoost models.
     """
 
     estimator_params = ClassifierDecisionTree.estimator_params + ["nb_features"]

art/estimators/classification/blackbox.py

@@ -284,7 +284,7 @@ def loss_gradient(  # pylint: disable=W0221
     def layer_names(self) -> List[str]:
         """
         Return the hidden layers in the model, if applicable. This function is not supported for the
-        Classifier and Detector wrapper.
+        Classifier and Detector classes.
 
         :return: The hidden layers in the model, input and output layers excluded.
         :raises `NotImplementedException`: This method is not supported for detector-classifiers.

art/estimators/classification/catboost.py

@@ -45,7 +45,7 @@
 
 class LightGBMClassifier(ClassifierDecisionTree):
     """
-    Wrapper class for importing LightGBM models.
+    Class for importing LightGBM models.
     """
 
     def __init__(

art/estimators/classification/detector_classifier.py

@@ -46,7 +46,7 @@
 
 class MXClassifier(ClassGradientsMixin, ClassifierMixin, MXEstimator):  # lgtm [py/missing-call-to-init]
     """
-    Wrapper class for importing MXNet Gluon models.
+    Class for importing MXNet Gluon models.
     """
 
     estimator_params = (

art/estimators/classification/lightgbm.py

@@ -18,38 +18,31 @@
 """
 Provides black-box gradient estimation using NES.
 """
-from __future__ import absolute_import, division, print_function, unicode_literals
-
 import logging
 from typing import List, Optional, Tuple, Union, TYPE_CHECKING
 
 import numpy as np
 from scipy.stats import entropy
 
-from art.estimators.classification.classifier import ClassifierClassLossGradients
-from art.utils import clip_and_round, deprecated
-from art.wrappers.wrapper import ClassifierWrapper
+from art.estimators.estimator import BaseEstimator
+from art.estimators.classification.classifier import ClassifierMixin, ClassifierLossGradients
+from art.utils import clip_and_round
 
 if TYPE_CHECKING:
     from art.utils import CLASSIFIER_CLASS_LOSS_GRADIENTS_TYPE
 
 logger = logging.getLogger(__name__)
 
 
-@deprecated(
-    end_version="1.8.0",
-    reason="Expectation over transformation has been replaced with " "art.estimators",
-    replaced_by="art.preprocessing.expectation_over_transformation",
-)
-class QueryEfficientBBGradientEstimation(ClassifierWrapper, ClassifierClassLossGradients):
+class QueryEfficientGradientEstimationClassifier(ClassifierLossGradients, ClassifierMixin, BaseEstimator):
     """
     Implementation of Query-Efficient Black-box Adversarial Examples. The attack approximates the gradient by
     maximizing the loss function over samples drawn from random Gaussian noise around the input.
 
     | Paper link: https://arxiv.org/abs/1712.07113
     """
 
-    attack_params = ["num_basis", "sigma", "round_samples"]
+    estimator_params = ["num_basis", "sigma", "round_samples"]
 
     def __init__(
         self,
@@ -59,20 +52,19 @@ def __init__(
         round_samples: float = 0.0,
     ) -> None:
         """
-        :param classifier: An instance of a `Classifier` whose loss_gradient is being approximated.
+        :param classifier: An instance of a classification estimator whose loss_gradient is being approximated.
         :param num_basis:  The number of samples to draw to approximate the gradient.
         :param sigma: Scaling on the Gaussian noise N(0,1).
         :param round_samples: The resolution of the input domain to round the data to, e.g., 1.0, or 1/255. Set to 0 to
                               disable.
         """
-        super().__init__(classifier)
-        # self.predict refers to predict of classifier
+        super().__init__(model=classifier.model, clip_values=classifier.clip_values)
         # pylint: disable=E0203
-        self._predict = self.classifier.predict
+        self._classifier = classifier
         self.num_basis = num_basis
         self.sigma = sigma
         self.round_samples = round_samples
-        self._nb_classes = self.classifier.nb_classes
+        self._nb_classes = self._classifier.nb_classes
 
     @property
     def input_shape(self) -> Tuple[int, ...]:
@@ -81,18 +73,18 @@ def input_shape(self) -> Tuple[int, ...]:
 
         :return: Shape of one input sample.
         """
-        return self._input_shape  # type: ignore
+        return self._classifier.input_shape  # type: ignore
 
     def predict(self, x: np.ndarray, batch_size: int = 128, **kwargs) -> np.ndarray:  # pylint: disable=W0221
         """
-        Perform prediction of the classifier for input `x`.
+        Perform prediction of the classifier for input `x`. Rounds results first.
 
         :param x: Features in array of shape (nb_samples, nb_features) or (nb_samples, nb_pixels_1, nb_pixels_2,
                   nb_channels) or (nb_samples, nb_channels, nb_pixels_1, nb_pixels_2).
         :param batch_size: Size of batches.
         :return: Array of predictions of shape `(nb_inputs, nb_classes)`.
         """
-        return self._wrap_predict(x, batch_size=batch_size)
+        return self._classifier.predict(clip_and_round(x, self.clip_values, self.round_samples), batch_size=batch_size)
 
     def fit(self, x: np.ndarray, y: np.ndarray, **kwargs) -> None:
         """
@@ -172,16 +164,6 @@ def loss_gradient(self, x: np.ndarray, y: np.ndarray, **kwargs) -> np.ndarray:
         grads = self._apply_preprocessing_gradient(x, np.array(grads))
         return grads
 
-    def _wrap_predict(self, x: np.ndarray, batch_size: int = 128) -> np.ndarray:
-        """
-        Perform prediction for a batch of inputs. Rounds results first.
-
-        :param x: Input samples.
-        :param batch_size: Size of batches.
-        :return: Array of predictions of shape `(nb_inputs, nb_classes)`.
-        """
-        return self._predict(clip_and_round(x, self.clip_values, self.round_samples), **{"batch_size": batch_size})
-
     def get_activations(self, x: np.ndarray, layer: Union[int, str], batch_size: int) -> np.ndarray:
         """
         Return the output of the specified layer for input `x`. `layer` is specified by layer index (between 0 and

art/estimators/classification/mxnet.py

@@ -62,7 +62,7 @@ def SklearnClassifier(
 ) -> "ScikitlearnClassifier":
     """
     Create a `Classifier` instance from a scikit-learn Classifier model. This is a convenience function that
-    instantiates the correct wrapper class for the given scikit-learn model.
+    instantiates the correct class for the given scikit-learn model.
 
     :param model: scikit-learn Classifier model.
     :param clip_values: Tuple of the form `(min, max)` representing the minimum and maximum values allowed
@@ -100,7 +100,7 @@ def SklearnClassifier(
 
 class ScikitlearnClassifier(ClassifierMixin, ScikitlearnEstimator):  # lgtm [py/missing-call-to-init]
     """
-    Wrapper class for scikit-learn classifier models.
+    Class for scikit-learn classifier models.
     """
 
     estimator_params = ClassifierMixin.estimator_params + ScikitlearnEstimator.estimator_params + ["use_logits"]
@@ -284,7 +284,7 @@ def _get_nb_classes(self) -> int:
 
 class ScikitlearnDecisionTreeClassifier(ScikitlearnClassifier):
     """
-    Wrapper class for scikit-learn Decision Tree Classifier models.
+    Class for scikit-learn Decision Tree Classifier models.
     """
 
     def __init__(
@@ -430,7 +430,7 @@ def _get_leaf_nodes(self, node_id, i_tree, class_label, box) -> List["LeafNode"]
 
 class ScikitlearnDecisionTreeRegressor(ScikitlearnDecisionTreeClassifier):
     """
-    Wrapper class for scikit-learn Decision Tree Regressor models.
+    Class for scikit-learn Decision Tree Regressor models.
     """
 
     def __init__(
@@ -520,7 +520,7 @@ def _get_leaf_nodes(self, node_id, i_tree, class_label, box) -> List["LeafNode"]
 
 class ScikitlearnExtraTreeClassifier(ScikitlearnDecisionTreeClassifier):
     """
-    Wrapper class for scikit-learn Extra TreeClassifier Classifier models.
+    Class for scikit-learn Extra TreeClassifier Classifier models.
     """
 
     def __init__(
@@ -559,7 +559,7 @@ def __init__(
 
 class ScikitlearnAdaBoostClassifier(ScikitlearnClassifier):
     """
-    Wrapper class for scikit-learn AdaBoost Classifier models.
+    Class for scikit-learn AdaBoost Classifier models.
     """
 
     def __init__(
@@ -598,7 +598,7 @@ def __init__(
 
 class ScikitlearnBaggingClassifier(ScikitlearnClassifier):
     """
-    Wrapper class for scikit-learn Bagging Classifier models.
+    Class for scikit-learn Bagging Classifier models.
     """
 
     def __init__(
@@ -638,7 +638,7 @@ def __init__(
 
 class ScikitlearnExtraTreesClassifier(ScikitlearnClassifier, DecisionTreeMixin):
     """
-    Wrapper class for scikit-learn Extra Trees Classifier models.
+    Class for scikit-learn Extra Trees Classifier models.
     """
 
     def __init__(
@@ -711,7 +711,7 @@ def get_trees(self) -> List["Tree"]:  # lgtm [py/similar-function]
 
 class ScikitlearnGradientBoostingClassifier(ScikitlearnClassifier, DecisionTreeMixin):
     """
-    Wrapper class for scikit-learn Gradient Boosting Classifier models.
+    Class for scikit-learn Gradient Boosting Classifier models.
     """
 
     def __init__(
@@ -785,7 +785,7 @@ def get_trees(self) -> List["Tree"]:
 
 class ScikitlearnRandomForestClassifier(ScikitlearnClassifier):
     """
-    Wrapper class for scikit-learn Random Forest Classifier models.
+    Class for scikit-learn Random Forest Classifier models.
     """
 
     def __init__(
@@ -858,7 +858,7 @@ def get_trees(self) -> List["Tree"]:  # lgtm [py/similar-function]
 
 class ScikitlearnLogisticRegression(ClassGradientsMixin, LossGradientsMixin, ScikitlearnClassifier):
     """
-    Wrapper class for scikit-learn Logistic Regression models.
+    Class for scikit-learn Logistic Regression models.
     """
 
     def __init__(
@@ -1045,7 +1045,7 @@ def get_trainable_attribute_names() -> Tuple[str, str]:
 
 class ScikitlearnGaussianNB(ScikitlearnClassifier):
     """
-    Wrapper class for scikit-learn Gaussian Naive Bayes models.
+    Class for scikit-learn Gaussian Naive Bayes models.
     """
 
     def __init__(
@@ -1094,7 +1094,7 @@ def get_trainable_attribute_names() -> Tuple[str, str]:
 
 class ScikitlearnSVC(ClassGradientsMixin, LossGradientsMixin, ScikitlearnClassifier):
     """
-    Wrapper class for scikit-learn C-Support Vector Classification models.
+    Class for scikit-learn C-Support Vector Classification models.
     """
 
     def __init__(