-
I use the gaussian augmentation and label smoothing method on the cifar-10 adversarial training example. But the accuracy of defences methods is lower than the clean train dataset. Is this normal?
-
There is an error when we initialize the class InceptionResnetV1Model.
`super(InceptionResnetV1Model, self).__init__('model', None, {})`
It will show
`TypeError: __init__() takes exactly 1 argument…
-
Hi,
I'm trying to use fgsm on RandomForest but I couldnt do it. Details are [here](
https://stackoverflow.com/questions/53076989/randomforestclassifier-object-has-no-attribute-layers). Then I saw …
-
Thank you for replying our question so quickly and your suggestions are really useful!
But our scale for pixel is somehow abnormal which is (-255,0).
```python
# Attack
with foolbox.models.Tens…
-
In
https://github.com/cihangxie/DI-2-FGSM/blob/dfa64f2842561518f99515ea9572b42f9bee0c62/attack.py#L179
the variable `grad` is initialized to zero. Presumably, `grad` stores the accumulated gradie…
-
I'm trying to generate the Table 4, Row 1 results of the paper.
Set up is a white box, Model A, FGSM attack 0.3, No defense, 50 epochs, 1e-3 lr, MNIST, Adam optimizer. While the paper reports 99.…
-
您好!
最近在做Keras下的FGSM实验，发现KerasModel中有FeatureFqueezingDefence这个函数。
请问，这是表示默认情况下都会加入特征挤压防御嘛？如果只是想得到原始的模型，在KerasModel初始化时，把featurefqueezing_bit_depth设置为None，则scaled_data会跳过下列if语句：
`if self._feat…
-
Hi, it seems that some functions miss required positional arguments, e.g.,
in test_dense_cnn.py line 81, model.get_logits(x) misses 'reuse',
in test_dense_cnn.py line 86, x_fgsm = fgsm.generate(x, e…
-
I'd like to view the maximized crossentropy after the attack instead of accuracy, say with FGSM. How can I do that simply?
-
Hi,
In order to evaluate an adversarial sample detection mechanism, I was looking for an option to get the failed adversarial sample for a given attack.
I tried redefining my own Criterion to a…