PGD:Towards Deep Learning Models Resistant to Adversarial Attacks
一.摘要 1.出发点：（1）深度网络对对抗攻击（adversarial attack）即对对抗样本的天生缺陷。 （2）当输入信息与自然样本差别并不大是却被网络错误的分类，如下图。 2.作者提出方法和论证如下： （1）从鲁棒优化的角度研究了神经网络的抗差鲁棒性。这与很多最近的相关工作具有广泛的一致性 eg：防御蒸馏，特征压缩。 （2）根据方法的原理性使得该方法在普遍范围内是通用和有效的，即使用最强的
相关文章
- 1. 对抗机器学习——Min Max模型(Towards Deep Learning Models Resistant to Adversarial Attacks)
- 2. [论文解读]Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey
- 3. [转载][paper]Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey
- 4. Model-Reuse Attacks on Deep Learning Systems
- 5. 关于EAD: Elastic-Net Attacks to Deep Neural Networks via Adversarial Examples的理解
- 6. Deep Learning Models on Kubernetes with GPUs
- 7. Deploying deep learning models in OpenCV
- 8. 【对抗攻击】【综述】Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey
- 9. How To Improve Deep Learning Performance
- 10. WHEN NOT TO USE DEEP LEARNING
- 更多相关文章...
- • XQuery 选择 和 过滤 - XQuery 教程
- • ASP.NET MVC - 模型 - ASP.NET 教程
- • Java Agent入门实战（一）-Instrumentation介绍与使用
- • Java Agent入门实战（三）-JVM Attach原理与使用
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. 对抗机器学习——Min Max模型(Towards Deep Learning Models Resistant to Adversarial Attacks)
- 2. [论文解读]Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey
- 3. [转载][paper]Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey
- 4. Model-Reuse Attacks on Deep Learning Systems
- 5. 关于EAD: Elastic-Net Attacks to Deep Neural Networks via Adversarial Examples的理解
- 6. Deep Learning Models on Kubernetes with GPUs
- 7. Deploying deep learning models in OpenCV
- 8. 【对抗攻击】【综述】Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey
- 9. How To Improve Deep Learning Performance
- 10. WHEN NOT TO USE DEEP LEARNING