ImageNet Pre-training Also Transfers Non-robustness
Jiaming Zhang, Jitao Sang, Qi Yi, Yunfan Yang, Huiwen Dong, Jian Yu
Abstract
ImageNet pre-training has enabled state-of-the-art results on many tasks. In spite of its recognized contribution to generalization, we observed in this study that ImageNet pre-training also transfers adversarial non-robustness from pre-trained model into fine-tuned model in the downstream classification tasks. We first conducted experiments on various datasets and network backbones to uncover the adversarial non-robustness in fine-tuned model. Further analysis was conducted on examining the learned knowledge of fine-tuned model and standard model, and revealed that the reason leading to the non-robustness is the non-robust features transferred from ImageNet pre-trained model. Finally, we analyzed the preference for feature learning of the pre-trained model, explored the factors influencing robustness, and introduced a simple robust ImageNet pre-training solution. Our code is available at https://github.com/jiamingzhang94/ImageNet-Pretraining-transfers-non-robustness.
BibTeX
@article{Zhang_Sang_Yi_Yang_Dong_Yu_2023, title={ImageNet Pre-training Also Transfers Non-robustness}, volume={37}, url={https://ojs.aaai.org/index.php/AAAI/article/view/25452}, DOI={10.1609/aaai.v37i3.25452}, abstractNote={ImageNet pre-training has enabled state-of-the-art results on many tasks. In spite of its recognized contribution to generalization, we observed in this study that ImageNet pre-training also transfers adversarial non-robustness from pre-trained model into fine-tuned model in the downstream classification tasks. We first conducted experiments on various datasets and network backbones to uncover the adversarial non-robustness in fine-tuned model. Further analysis was conducted on examining the learned knowledge of fine-tuned model and standard model, and revealed that the reason leading to the non-robustness is the non-robust features transferred from ImageNet pre-trained model. Finally, we analyzed the preference for feature learning of the pre-trained model, explored the factors influencing robustness, and introduced a simple robust ImageNet pre-training solution. Our code is available at https://github.com/jiamingzhang94/ImageNet-Pretraining-transfers-non-robustness.}, number={3}, journal={Proceedings of the AAAI Conference on Artificial Intelligence}, author={Zhang, Jiaming and Sang, Jitao and Yi, Qi and Yang, Yunfan and Dong, Huiwen and Yu, Jian}, year={2023}, month={Jun.}, pages={3436-3444} }