 |
ML Newsnew | past | best | rss | submit | about |
| Explaining and Harnessing Adversarial Examples [problem with deep networks] (arxiv.org) | |
| 1 point by alexanderg on Dec 23, 2014 | hide | past | pdf | 1 comment on HN | |
|
In plain words: Tiny worst-case changes to an image fool neural networks into wrong answers; the cause is their linear behavior, not overfitting. This explains why the trick fools other networks, and gives a fast way to make examples whose use in training lowers error on handwritten digits. Abstract · Explaining and Harnessing Adversarial ExamplesSeveral machine learning models, including neural networks, consistently misclassify adversarial examples---inputs formed by applying small but intentionally worst-case perturbations to examples from the dataset, such that the perturbed input results in the model outputting an incorrect answer with high confidence. Early attempts at explaining this phenomenon focused on nonlinearity and overfitting. We argue instead that the primary cause of neural networks' vulnerability to adversarial perturbation is their linear nature. This explanation is supported by new quantitative results while giving the first explanation of the most intriguing fact about them: their generalization across architectures and training sets. Moreover, this view yields a simple and fast method of generating adversarial examples. Using this approach to provide examples for adversarial training, we reduce the test set error of a maxout network on the MNIST dataset. | |
| add comment on HN |
These results suggest that classifiers based on modern machine learning techniques, even those that obtain excellent performance on the test set, are not learning the true underlying concepts that determine the correct output label. Instead, these algorithms have built a Potemkin village that works well on naturally occuring data, but is exposed as a fake when one visits points in space that do not have high probability in the data distribution. This is particularly disappointing because a popular approach in computer vision is to use convolutional network features as a space where Euclidean distance approximates perceptual distance. This resemblance is clearly flawed if images that have an immeasurably small perceptual distance correspond to completely different classes in the network’s representation.
These results have often been interpreted as being a flaw in deep networks in particular, even though linear classifiers have the same problem. We regard the knowledge of this flaw as an opportunity to fix it. Indeed, Gu & Rigazio (2014) and Chalupka et al. (2014) have already begun the first steps toward designing models that resist adversarial perturbation, though no model has yet succesfully done so while maintaining state of the art accuracy on clean inputs.