MART

Code for ICLR2020 "Improving Adversarial Robustness Requires Revisiting Misclassified Examples"

Usage

python3 train_resnet.py for ResNet18

python3 train_wideresnet.py for WideResNet

Trained Models

The ResNet18 trained by MART on CIFAR-10: https://drive.google.com/file/d/1YAKnAhUAiv8UFHnZfj2OIHWHpw_HU0Ig/view?usp=sharing

The WideResNet-34-10 trained by MART on CIFAR-10: https://drive.google.com/open?id=1QjEwSskuq7yq86kRKNv6tkn9I16cEBjc

MART WideResNet-28-10 model on 500K unlabeled data: https://drive.google.com/file/d/11pFwGmLfbLHB4EvccFcyHKvGb3fBy_VY/view?usp=sharing

Citing this work

If you use this code in your work, please cite the accompanying paper:

@inproceedings{Wang2020Improving,
    title={Improving Adversarial Robustness Requires Revisiting Misclassified Examples},
    author={Yisen Wang and Difan Zou and Jinfeng Yi and James Bailey and Xingjun Ma and Quanquan Gu},
    booktitle={ICLR},
    year={2020}
}

Requirements

• Python 3.7.4
• Pytorch 1.3.1
• Part of the code is based on the following repo:
  • Dynamic: https://github.com/YisenWang/dynamic_adv_training
  • TREADES: https://github.com/yaodongyu/TRADES
  • RST: https://github.com/yaircarmon/semisup-adv