Clarifying the Behavior and the Difficulty of Adversarial Training

Authors: Xu Cheng, Hao Zhang, Yue Xin, Wen Shen, Quanshi Zhang

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Despite the simplicity of our theory, it still reveals verifiable predictions about various phenomena in adversarial training under real-world settings. Experimental verification 1 of Theorem 2. Although Theorem 2 was derived on a simple two-layer network, we tested whether our theory could predict the dynamics of adversarial perturbations on deep networks. That is, we checked whether ˆδ derived in Theorem 2 fitted well with the real perturbation δ .
Researcher Affiliation Academia Xu Cheng1,2, Hao Zhang2, Yue Xin2, Wen Shen2, Quanshi Zhang2* 1Nanjing University of Science and Technology, 2Shanghai Jiao Tong University
Pseudocode No The paper presents mathematical formulations and theorems but does not include any pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any explicit statements about releasing source code or provide links to a code repository.
Open Datasets Yes To this end, we crafted perturbations δ on different Re LU networks with more than two linear layers for the MNIST dataset (Le Cun et al. 1998).
Dataset Splits No The paper mentions using the MNIST dataset but does not provide specific details on how the dataset was split into training, validation, and test sets, or reference standard splits.
Hardware Specification No The paper does not specify any hardware details (e.g., specific GPU/CPU models, memory, or cloud instances) used for running the experiments.
Software Dependencies No The paper mentions using various deep learning models (e.g., VGG11, Alex Net, Res Net-18) but does not list specific software dependencies with version numbers (e.g., PyTorch 1.9, Python 3.8).
Experiment Setup Yes Please see Appendix L for the hyper-parameters of the attack.