On Connections Between Regularizations for Improving DNN Robustness
IEEE Transactions on Pattern Analysis and Machine Intelligence(2021)
摘要
This paper analyzes regularization terms proposed recently for improving the adversarial robustness of deep neural networks (DNNs), from a theoretical point of view. Specifically, we study possible connections between several effective methods, including input-gradient regularization, Jacobian regularization, curvature regularization, and a cross-Lipschitz functional. We investigate them on DNNs w...
更多查看译文
关键词
Robustness,Jacobian matrices,Training,Perturbation methods,Neural networks,Computational modeling,Task analysis
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要