You Only Propagate Once: Accelerating Adversarial Training via Maximal Principle

Authors: Dinghuai Zhang, Tianyuan Zhang, Yiping Lu, Zhanxing Zhu, Bin Dong

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical experiments demonstrate that YOPO can achieve comparable defense accuracy with approximately 1/5 1/4 GPU time of the projected gradient descent (PGD) algorithm.
Researcher Affiliation Academia Dinghuai Zhang , Tianyuan Zhang Peking University {zhangdinghuai, 1600012888}@pku.edu.cn; Yiping Lu Stanford University yplu@stanford.edu; Zhanxing Zhu School of Mathematical Sciences, Peking University Center for Data Science, Peking University Beijing Institute of Big Data Research zhanxing.zhu@pku.edu.cn; Bin Dong Beijing International Center for Mathematical Research, Peking University Center for Data Science, Peking University Beijing Institute of Big Data Research dongbin@math.pku.edu.cn
Pseudocode Yes Algorithm 1 YOPO (You Only Propagate Once)
Open Source Code Yes Our codes are available at https://github.com/a1600012888/YOPO-You-Only-Propagate-Once
Open Datasets Yes To demonstrate the effectiveness of YOPO, we conduct experiments on MNIST and CIFAR10.
Dataset Splits No The paper mentions using MNIST and CIFAR10 but does not specify the training, validation, and test dataset splits (e.g., percentages or sample counts).
Hardware Specification No The paper mentions "GPU time" but does not specify the type or model of GPUs, CPUs, or any other hardware components used for experiments.
Software Dependencies No The paper does not list any specific software dependencies with their version numbers (e.g., Python 3.x, PyTorch 1.x).
Experiment Setup Yes As a comparison, we test YOPO-3-5 and YOPO-5-3 with a step size of 2/255.