Update example scripts

examples/adversarial_training_FBF.py

@@ -11,7 +11,7 @@
 import torchvision.transforms as transforms
 from torch.utils.data import Dataset, DataLoader
 
-from art.classifiers import PyTorchClassifier
+from art.estimators.classification import PyTorchClassifier
 from art.data_generators import PyTorchDataGenerator
 from art.defences.trainer import AdversarialTrainerFBFPyTorch
 from art.utils import load_cifar10

examples/adversarial_training_data_augmentation.py

@@ -1,6 +1,10 @@
 """
 This is an example of how to use ART and Keras to perform adversarial training using data generators for CIFAR10
 """
+import tensorflow as tf
+
+tf.compat.v1.disable_eager_execution()
+
 import keras
 import numpy as np
 from keras.layers import Conv2D, Dense, Flatten, MaxPooling2D, Input, BatchNormalization

examples/get_started_lightgbm.py

@@ -27,7 +27,7 @@
 
 # Step 2: Create the model
 
-params = {"objective": "multiclass", "metric": "multi_logloss", "num_class": 10}
+params = {"objective": "multiclass", "metric": "multi_logloss", "num_class": 10, "force_col_wise": True}
 train_set = lgb.Dataset(x_train, label=np.argmax(y_train, axis=1))
 test_set = lgb.Dataset(x_test, label=np.argmax(y_test, axis=1))
 model = lgb.train(params=params, train_set=train_set, num_boost_round=100, valid_sets=[test_set])

examples/get_started_xgboost.py

@@ -27,7 +27,7 @@
 
 # Step 2: Create the model
 
-params = {"objective": "multi:softprob", "metric": "accuracy", "num_class": 10}
+params = {"objective": "multi:softprob", "eval_metric": ["mlogloss", "merror"], "num_class": 10}
 dtrain = xgb.DMatrix(x_train, label=np.argmax(y_train, axis=1))
 dtest = xgb.DMatrix(x_test, label=np.argmax(y_test, axis=1))
 evals = [(dtest, "test"), (dtrain, "train")]

examples/mnist_cnn_fgsm.py

@@ -2,6 +2,10 @@
 """Trains a convolutional neural network on the MNIST dataset, then attacks it with the FGSM attack."""
 from __future__ import absolute_import, division, print_function, unicode_literals
 
+import tensorflow as tf
+
+tf.compat.v1.disable_eager_execution()
+
 from keras.models import Sequential
 from keras.layers import Dense, Flatten, Conv2D, MaxPooling2D, Dropout
 import numpy as np
@@ -35,12 +39,16 @@
 acc = np.sum(preds == np.argmax(y_test, axis=1)) / y_test.shape[0]
 print("\nTest accuracy: %.2f%%" % (acc * 100))
 
-# Craft adversarial samples with FGSM
-epsilon = 0.1  # Maximum perturbation
-adv_crafter = FastGradientMethod(classifier, eps=epsilon)
-x_test_adv = adv_crafter.generate(x=x_test)
+# Define epsilon values
+epsilon_values = [0.01, 0.1, 0.15, 0.2, 0.25, 0.3]
 
-# Evaluate the classifier on the adversarial examples
-preds = np.argmax(classifier.predict(x_test_adv), axis=1)
-acc = np.sum(preds == np.argmax(y_test, axis=1)) / y_test.shape[0]
-print("\nTest accuracy on adversarial sample: %.2f%%" % (acc * 100))
+# Iterate over epsilon values
+for epsilon in epsilon_values:
+    # Craft adversarial samples with FGSM
+    adv_crafter = FastGradientMethod(classifier, eps=epsilon)
+    x_test_adv = adv_crafter.generate(x=x_test, y=y_test)
+
+    # Evaluate the classifier on the adversarial examples
+    preds = np.argmax(classifier.predict(x_test_adv), axis=1)
+    acc = np.sum(preds == np.argmax(y_test, axis=1)) / y_test.shape[0]
+    print("Test accuracy on adversarial sample (epsilon = %.2f): %.2f%%" % (epsilon, acc * 100))

examples/mnist_poison_detection.py

@@ -5,7 +5,9 @@
 
 import pprint
 import json
+import tensorflow as tf
 
+tf.compat.v1.disable_eager_execution()
 from keras.models import Sequential
 from keras.layers import Dense, Flatten, Conv2D, MaxPooling2D, Dropout
 import numpy as np

examples/mnist_transferability.py

@@ -12,6 +12,8 @@
 import numpy as np
 import tensorflow as tf
 
+tf.compat.v1.disable_eager_execution()
+
 from art.attacks.evasion import DeepFool
 from art.estimators.classification import KerasClassifier, TensorFlowClassifier
 from art.utils import load_mnist
@@ -60,7 +62,7 @@ def cnn_mnist_k(input_shape):
 
 
 # Get session
-session = tf.Session()
+session = tf.compat.v1.Session()
 k.set_session(session)
 
 # Read MNIST dataset