[paper]DeepFool: a simple and accurate method to fool deep neural networks
本文目标是寻求最小的扰动来达到生成对抗样本的目标，因此提出了DeepFool的算法来生成扰动，并且提出了一种量化分类器鲁棒性的方法。 这是第一个通过计算出最小的必要扰动，并应用到对抗样本构建的方法，使用的限制扰动规模的方法是L2范数。最终得到的对抗样本效果要优于前面的FGSM和JSMA方法，但是这三者都需要比较大的计算资源。 根据样本 x x x，标签 k ^ ( x ) \hat{k}(x) k
相关文章
- 1. DeepFool: a simple and accurate method to fool deep neural networks
- 2. 【论文快读】DeepFool: a simple and accurate method to fool deep neural networks(2016)
- 3. Pseudo-Label：The Simple and Efficient Semi-Supervised Learning Method for Deep Neural Networks
- 4. Neural Networks and Deep Learning -- Class 4: Deep Neural Networks
- 5. Neural Networks and Deep Learning
- 6. Neural Networks and Deep Learning -- Class 3: Shallow neural networks
- 7. Neural Networks and Deep Learning_week2练习
- 8. COURSE 1 Neural Networks and Deep Learning
- 9. Neural Networks and Deep Learning-引论
- 10. neural networks and deep learning-chapter1-3
- 更多相关文章...
- • XSL-FO simple-page-master 对象 - XSL-FO 教程
- • XSL-FO table-and-caption 对象 - XSL-FO 教程
- • RxJava操作符（七）Conditional and Boolean
- • Kotlin学习（二）基本类型
相关标签/搜索
每日一句
-
每一个你不满意的现在，都有一个你没有努力的曾经。
欢迎关注本站公众号,获取更多信息
相关文章
- 1. DeepFool: a simple and accurate method to fool deep neural networks
- 2. 【论文快读】DeepFool: a simple and accurate method to fool deep neural networks(2016)
- 3. Pseudo-Label：The Simple and Efficient Semi-Supervised Learning Method for Deep Neural Networks
- 4. Neural Networks and Deep Learning -- Class 4: Deep Neural Networks
- 5. Neural Networks and Deep Learning
- 6. Neural Networks and Deep Learning -- Class 3: Shallow neural networks
- 7. Neural Networks and Deep Learning_week2练习
- 8. COURSE 1 Neural Networks and Deep Learning
- 9. Neural Networks and Deep Learning-引论
- 10. neural networks and deep learning-chapter1-3